CN114066857A - Infrared image quality evaluation method and device, electronic equipment and readable storage medium - Google Patents

Abstract

The present application discloses an infrared image quality evaluation method, device, electronic device and readable storage medium. Among them, the method includes generating a training set and a test set with images carrying subjective image quality scores based on the original infrared image data set; training a machine learning model by using each training image of the training set and its subjective image quality scores to obtain an initial quality evaluation model; The no-reference evaluation algorithm calculates the objective score of each objective quality index of each test image in the test set, and determines the objective sub-dimension evaluation score of each test image; input each objective sub-dimension evaluation score into the initial quality evaluation model to obtain each test image Based on the initial quality evaluation model, the final image quality evaluation model is determined according to the image quality objective score of each test image and the corresponding image quality subjective score, so that it can efficiently and conveniently achieve a high degree of consistency with the subjective perception of the human eye. Infrared image quality evaluation.

Description

Infrared image quality evaluation method, device, electronic device and readable storage medium

technical field

The present application relates to the technical field of image processing, and in particular, to an infrared image quality evaluation method, device, electronic device, and readable storage medium.

Background technique

It is well known that, different from the principle of visible light imaging, the infrared thermal imaging system performs imaging by sensing the temperature difference between the thermal radiation of the object and the background thermal radiation. Due to its passive imaging characteristics, infrared imaging equipment can clearly capture images in dark nights, scenes with poor lighting environment, or harsh environments with dense smoke and dust, such as rainy and foggy days, hazy days, etc., which makes up for the fact that visible light imaging is limited by illumination Condition defects have been widely used in military, industrial, automotive assisted driving, security, medicine and other fields.

Due to the long imaging wavelength, infrared images generally suffer from large noise, low image contrast, low signal-to-noise ratio, unclear edges, blurred visual effects, and narrow gray scale. In the process of image post-processing, some noise, blur, and even some information are inevitably introduced. These factors will lead to the degradation of image quality or image distortion. The quality of infrared images directly determines the user's visual experience and the acquisition of information, and the evaluation of infrared image quality is particularly important. Although there are parameters such as the minimum distinguishable temperature difference and the noise equivalent temperature difference in infrared imaging, these indicators cannot objectively reflect the user's subjective feeling of the image. Therefore, a general-purpose infrared image quality evaluation method that is highly consistent with the user's subjective experience was developed. Through this method, the user can accurately know the quality of the infrared image to be evaluated, and then use it to guide the construction and adjustment of the infrared image acquisition equipment and processing system. , as well as optimizing image processing algorithms and parameter settings to finally present higher-quality images to users.

In the visible light field, the current image quality evaluation methods include subjective evaluation method and objective evaluation method. The subjective evaluation method is that the observer subjectively scores the image, and then calculates the average subjective score or the difference value of the average subjective score. Specifically, it can be divided into absolute evaluation method and relative evaluation method. The objective evaluation method is that the computer obtains the quality index of the image according to a certain algorithm. According to whether the reference image needs to be introduced in the evaluation process, it is divided into full reference, semi-reference and no-reference evaluation methods. The non-reference evaluation method is also called blind image quality. Evaluation. In the field of infrared imaging, the subjective evaluation method is still applicable. The advantage of this method is that it is accurate and reliable. The disadvantage is that it is subject to the influence of the observer's professional background, psychology, motivation and other factors. Time consuming and laborious. As for the objective evaluation method, the full reference evaluation model often has the best effect because it can use all image information. However, in the process of infrared imaging, since the original image without distortion cannot be obtained, the objective evaluation of infrared images can only be With the help of a no-reference evaluation method. The no-reference evaluation method is divided into the evaluation for specific distortion and the evaluation for non-specific distortion types. The evaluation method for specific distortion is more mature, and the most widely used is the evaluation of image blur and noise. PhotographicExperts Group, Joint Photographic Experts Group) compression distortion evaluation, but infrared images are often damaged by various distortions in practical applications, so evaluation algorithms for specific distortions cannot reflect the overall image quality level. The evaluation of non-specific distortion types is closer to the user evaluation method and has more use value. Related technologies are usually based on support vector machine methods such as BRISQUE (No-Reference Image Quality Assessment in the Spatial Domain, spatial domain non-reference image quality evaluation) algorithm, based on Probabilistic model methods such as NIQE (Natural image quality evaluator, natural image quality evaluation) algorithm, in addition to dictionary-based methods and neural network-based methods for image evaluation, but the above methods are complex in calculation process and can be operated in real application scenarios Bad sex.

In view of this, how to efficiently and conveniently realize infrared image quality evaluation that is highly consistent with the subjective perception of the human eye is a technical problem that needs to be solved by those in the art.

SUMMARY OF THE INVENTION

The present application provides an infrared image quality evaluation method, device, electronic device and readable storage medium, which can efficiently and conveniently realize infrared image quality evaluation that is highly consistent with the subjective perception of the human eye.

In order to solve the above-mentioned technical problems, the embodiments of the present invention provide the following technical solutions:

One aspect of the embodiments of the present invention provides an infrared image quality evaluation method, including:

Generate training sets and test sets with images carrying subjective scores of image quality based on the original infrared image dataset;

By using each training image in the training set and the corresponding subjective image quality score to train the machine learning model, an initial quality evaluation model is obtained;

Determine the objective sub-dimension evaluation score of each test image by calculating the objective score of each objective quality index of each test image in the test set by using a no-reference evaluation algorithm;

Inputting each objective sub-dimension evaluation score into the initial quality evaluation model to obtain the image quality objective score of each test image;

Based on the initial quality evaluation model, the final image quality evaluation model is determined according to the objective image quality scores of each test image and the corresponding subjective image quality scores.

Optionally, generating a training set and a test set of subjective scores of image quality based on the original infrared image data set, including:

Obtain infrared images collected by multiple infrared thermal imaging devices in various types of application scenarios under different lighting environments to form the original infrared image data set; wherein, the manufacturer, packaging method and resolution of each infrared thermal imaging device different;

obtaining the subjective image quality score of each infrared image in the original infrared image dataset;

Dividing each infrared image in the original infrared image data set into multiple training images and multiple testing images based on a preset division ratio;

Generate a training set according to multiple training images and the subjective image quality scores corresponding to each training image;

A test set is generated according to multiple test images and the subjective image quality scores corresponding to each test image.

Optionally, the obtaining the subjective image quality score of each infrared image in the original infrared image data set includes:

Obtain the subjective evaluation scores of each subjective quality index of each infrared image by multiple experts, so as to obtain the subjective sub-dimension evaluation score of each infrared image;

Obtain the overall subjective evaluation scores of each infrared image by multiple experts and multiple ordinary users, and calculate the subjective overall evaluation score of each infrared image according to the preset expert weight coefficient and user weight coefficient;

The subjective image quality score of each infrared image is determined according to the subjective sub-dimension evaluation score and the subjective overall evaluation score of each infrared image.

Optionally, the initial quality evaluation model is obtained by using each training image in the training set and the corresponding subjective image quality score to train the machine learning model, including:

Pre-built support vector regression models;

Construct the score feature vector according to the subjective sub-dimension evaluation score of each training image;

The score feature vector and the subjective overall evaluation score corresponding to each training image are input into the support vector regression model for training to obtain the initial quality evaluation model.

Optionally, the objective score of each objective quality index of each test image in the test set is calculated by using a no-reference evaluation algorithm to determine the objective sub-dimension evaluation score of each test image, including:

The non-uniformity objective score is obtained by calculating the image space variance of each test image and normalizing it;

Determine the objective score of image noise by calculating the local normalized luminance coefficient of each test image;

Determine the objective score of image clarity by calculating the structural similarity between each test image and the corresponding reference image;

Calculate the image contrast objective score of each test image based on the image spatial frequency;

Calculate the objective score of the dynamic brightness range of each test image based on the maximum gray level of the image and the minimum gray level of the image;

For each test image, normalize the non-uniformity objective score, image noise objective score, image clarity objective score, image contrast objective score, and dynamic brightness range objective score of the current test image to obtain the current test image The objective sub-dimension evaluation score of .

Optionally, determining the objective score of image noise by calculating the local normalized luminance coefficient of each test image, including:

For each test image, calculate the local normalized luminance coefficient of the current test image, and fit the local normalized luminance coefficient through a generalized Gaussian model to obtain the fitting parameter mean and fitting parameter variance;

calculating the local normalized luminance coefficient neighborhood coefficients of the local normalized luminance coefficient in multiple directions, and fitting each local normalized luminance coefficient neighborhood coefficient by an asymmetric generalized Gaussian model to obtain a plurality of fitting parameters;

Extract multi-dimensional statistical features from the fitting parameter mean, the fitting parameter variance, and each fitting parameter at different scales, and input the multi-dimensional statistical features into the initial quality evaluation model to obtain the current test image. Image noise objective score.

Optionally, determining the objective image clarity score by calculating the structural similarity between each test image and the corresponding reference image, including:

For each test image, use a Gaussian smoothing filter to perform low-pass filtering on the current test image to obtain a corresponding current reference image;

respectively extracting the gradient information of the current test image and the current reference image and the edge information in the target direction;

Generate a test gradient image according to the gradient information and edge information of the current test image; generate a reference gradient image according to the gradient information and edge information of the current reference image;

Determine from the test gradient image a plurality of target image blocks that satisfy the preset gradient information conditions, and determine that each target image block corresponds to a target reference image block in the reference gradient image;

Calling the pre-built image structure similarity relationship formula to calculate the structure similarity between each target image block and the corresponding target reference image block; wherein, the image structure similarity relationship formula is based on the brightness comparison function, the contrast comparison function and the structure information. The comparison functions and their respective weight coefficients are determined;

The objective score of the image clarity of the current test image is determined according to the structural similarity of each target image block.

Optionally, calculating the objective image contrast score of each test image based on the image spatial frequency includes:

For each test image, use a preset size template to traverse the current test image according to the spatial frequency of the horizontal direction and the vertical direction, and obtain the spatial frequency matrix of the current test image;

Based on the spatial frequency matrix, normalize the image spatial frequency of each pixel point one by one to obtain a normalized image spatial frequency matrix;

Determine the contrast sensitivity weight matrix of the current test image according to the pre-built contrast sensitivity relationship and the normalized image spatial frequency matrix;

The objective image contrast score of the current test image is calculated according to the contrast sensitivity weight matrix, the image size, the maximum gray level and the minimum gray level of the current test image.

Optionally, based on the initial quality evaluation model, the final image quality evaluation model is determined according to the objective image quality scores of each test image and the corresponding subjective image quality scores, including:

For each test image, the performance measurement indicators of the objective image quality score and the subjective image quality score of the current test image are respectively calculated; wherein, the performance measurement indicators include Pearson linear correlation coefficient, Spearman rank correlation coefficient, Kendall rank One or more of correlation coefficient, root mean square error;

If the performance measurement index satisfies a preset performance condition, the initial quality evaluation model is used as the image quality evaluation model;

If the performance measurement index does not meet the preset performance condition, an instruction to optimize the initial quality evaluation model is generated, and the initial quality evaluation model is retrained until the preset performance condition is met.

Another aspect of the embodiments of the present invention also provides an infrared image quality evaluation method, including:

Obtaining an image quality evaluation model in advance by using the infrared image quality evaluation method described in any one of the preceding items;

Obtain the infrared image to be evaluated;

Determine the objective sub-dimension evaluation score of the infrared image to be evaluated by calculating the objective score of each objective quality index of the infrared image to be evaluated by using the no-reference evaluation algorithm;

The objective sub-dimension evaluation score is input into the image quality evaluation model to obtain the image quality evaluation score of the infrared image to be evaluated.

Another aspect of the embodiments of the present invention also provides an infrared image quality evaluation device, including:

A dataset construction module, which is used to generate a training set and a test set of images carrying subjective scores of image quality based on the original infrared image dataset;

A model training module for training a machine learning model by using each training image in the training set and the corresponding subjective image quality scores to obtain an initial quality evaluation model;

a sub-dimension score calculation module, configured to determine the objective sub-dimension evaluation score of each test image by calculating the objective score of each objective quality index of each test image in the test set by using a no-reference evaluation algorithm;

an objective evaluation module, which inputs the evaluation scores of each objective sub-dimension into the initial quality evaluation model, and obtains an objective image quality score of each test image;

The model determination module is configured to determine the final image quality evaluation model according to the objective image quality scores and corresponding subjective image quality scores of each test image based on the initial quality evaluation model.

Another aspect of the embodiments of the present invention provides an infrared image quality evaluation device, including:

A model building module, used for obtaining an image quality evaluation model in advance using the infrared image quality evaluation method described in any of the preceding items;

an image acquisition module for acquiring the infrared image to be evaluated;

an objective scoring module, configured to determine the objective sub-dimension evaluation score of the infrared image to be evaluated by calculating the objective score of each objective quality index of the infrared image to be evaluated by using a reference-free evaluation algorithm;

A quality evaluation module, configured to input the objective sub-dimension evaluation scores into the image quality evaluation model to obtain the image quality evaluation scores of the infrared images to be evaluated.

An embodiment of the present invention further provides an electronic device, including a processor, which is configured to implement the steps of the infrared image quality evaluation method described in any preceding item when executing the computer program stored in the memory.

The embodiments of the present invention finally provide a readable storage medium, where a computer program is stored on the readable storage medium, and when the computer program is executed by a processor, the steps of the infrared image quality evaluation method described in any preceding item are implemented. .

The advantage of the technical solution provided by the present application is that the evaluation score of the subjective overall perception of the image by the human eye is used as the feature vector training model, and for the infrared image to be tested, an objective evaluation method oriented to specific distortion that is highly consistent with the subjective perception of the human eye is selected. Carry out objective evaluation, obtain the objective score value of the infrared image to be tested in each sub-dimension, input the objective score vector of each sub-dimension into the trained model, obtain the objective total score of the infrared image to be tested, and compare the objectivity of the subjective score with the objective score. The effective combination of objective scores, the final obtained objective infrared image quality score is highly consistent with the subjective perception of the human eye, which is very suitable for application scenarios where the current infrared field imaging content is complex and diverse, and the original undistorted image cannot be obtained. After the model training is completed, in the entire image quality evaluation process, only the objective sub-dimension evaluation score of the infrared image to be evaluated needs to be calculated, and the infrared image quality evaluation result that is highly consistent with the subjective perception of the human eye can be obtained. The infrared image quality evaluation that is highly consistent with the subjective perception of the human eye has strong operability and better practicability.

In addition, the embodiments of the present invention also provide a corresponding implementation device, electronic device, and readable storage medium for the infrared image quality evaluation method, which further makes the method more practical. The device, electronic device, and readable storage medium have corresponding advantages.

It is to be understood that the foregoing general description and the following detailed description are exemplary only and do not limit the present disclosure.

Description of drawings

In order to illustrate the technical solutions of the embodiments of the present invention or related technologies more clearly, the following briefly introduces the accompanying drawings that are used in the description of the embodiments or related technologies. Obviously, the drawings in the following description are only the present invention. For some embodiments of the present invention, for those of ordinary skill in the art, other drawings can also be obtained from these drawings without any creative effort.

1 is a schematic flowchart of a method for evaluating infrared image quality according to an embodiment of the present invention;

2 is a schematic flowchart of another infrared image quality evaluation method provided by an embodiment of the present invention;

3 is a schematic flowchart of still another infrared image quality evaluation method provided by an embodiment of the present invention;

FIG. 4 is a structural diagram of a specific implementation manner of an infrared image quality evaluation device provided by an embodiment of the present invention;

FIG. 5 is a structural diagram of another specific implementation manner of an infrared image quality evaluation device provided by an embodiment of the present invention;

FIG. 6 is a structural diagram of a specific implementation manner of an electronic device provided by an embodiment of the present invention.

Detailed ways

In order to make those skilled in the art better understand the solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some, but not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

The terms "first", "second", "third", "fourth", etc. in the description and claims of the present application and the above drawings are used to distinguish different objects, rather than to describe specific order. Furthermore, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or elements is not limited to the listed steps or elements, but may include unlisted steps or elements.

After introducing the technical solutions of the embodiments of the present invention, various non-limiting implementations of the present application are described in detail below.

Referring first to FIG. 1, FIG. 1 is a schematic flowchart of an infrared image quality evaluation method provided by an embodiment of the present invention. The embodiment of the present invention may include the following contents:

S101: Based on the original infrared image dataset, generate a training set and a test set in which images carry subjective scores of image quality.

In this embodiment, the original infrared image data set includes multiple infrared images, each infrared image is obtained by an infrared thermal imaging device collecting a target scene, and the target scene can be indoor, outdoor, any lighting condition, any Time period, scene in any external environment. The subjective image quality score is the user's subjective perception of the quality of infrared images based on their own, and is the subjective perception of the human eye. The image quality can be subjectively scored for each infrared image of the original infrared image dataset or a specified part of the infrared image, and the obtained subjective image quality score can be used as the label of each infrared image, and then according to a certain ratio, such as 8:2 Divide each infrared image of the original infrared image dataset into a training set and a test set. In order to distinguish the infrared images in each dataset, in this embodiment, the infrared images in the original infrared image dataset are called infrared images, the infrared images included in the training set are called training images, and the infrared images included in the test set are called testing images image. The number of infrared images contained in the training set and the test set can be flexibly selected according to the actual application scenario. Of course, in order to expand the original infrared image dataset, each infrared image in the original infrared image dataset can also be flipped, cropped, and denoised. After various processing, the infrared images obtained from the processing are supplemented into the original image data set. Of course, it is also possible to first divide the original infrared image dataset into a training set and a test set according to a certain ratio, and then perform image analysis on each training image in the training set or the designated training image, each test image in the test set or the designated test image. Quality subjective ratings, none of which affect the implementation of this application.

S102: Train the machine learning model by using each training image in the training set and the corresponding subjective image quality scores to obtain an initial quality evaluation model.

The machine learning model in this embodiment can be any existing machine learning model, such as a support vector machine model, a support vector regression model, a convolutional neural network model, a forward multilayer perceptron, and so on. Those skilled in the art The corresponding machine learning model can be flexibly selected according to actual needs, and the initial quality evaluation model for image quality evaluation can be obtained by using each training image in the training set and its corresponding subjective image quality score as sample data to train the machine learning model. As for the process of using the sample data to train the machine learning model, you can refer to the model training process described in the related art based on the type of the machine learning model used, and details are not repeated here.

S103: Determine the objective sub-dimension evaluation score of each test image by calculating the objective score of each objective quality index of each test image in the test set by using a no-reference evaluation algorithm.

In this step, for each test image in the test set, a no-reference evaluation algorithm for specific dimension distortion is introduced to calculate the objective sub-dimension score. For each test image, the objective sub-dimension evaluation score is determined by the objective score corresponding to each objective quality index. The objective sub-dimension evaluation score can be a data set or a matrix, and each element is the objective score of each objective quality index. Score, the objective sub-dimension evaluation score can also be a score, which is obtained by adding the objective scores corresponding to each objective quality index, or by weighting the summation of the objective scores corresponding to each objective quality index. The weight factor of the quality index can be set in advance according to the actual application scenario, and each test image corresponds to a set of objective sub-dimension evaluation scores.

S104: Input the evaluation scores of each objective sub-dimension into the initial quality evaluation model to obtain an objective image quality score of each test image.

After the objective evaluation algorithm for specific dimension distortion in the previous step calculates the objective score of the image in each sub-dimension, it is input into the initial quality evaluation model trained in S102, and the image quality score is output after the initial quality model calculation and processing. , the output is the objective total score of the test image.

S105: Based on the initial quality evaluation model, determine a final image quality evaluation model according to the objective image quality scores and corresponding subjective image quality scores of each test image.

It is understandable that the technical problem to be solved by this application is to obtain an infrared image quality evaluation result that is highly consistent with the subjective perception of the human eye, so that the final infrared image quality evaluation result is closer to the human evaluation result, but the evaluation process is independent of the human eye. Human subjective factors, and the image quality evaluation result that is highly consistent with the subjective perception of the human eye is output by the quality evaluation model. Based on this, the present application determines whether the performance of the initial quality evaluation model obtained by training in S102 meets the requirements by judging whether the image quality results output by the initial quality evaluation model are highly consistent with the subjective perception of the human eye. If it meets the requirements, you can directly use S102 As the quality evaluation model used in the actual infrared image evaluation process, if it does not meet the requirements, the initial quality evaluation model needs to be optimized until it meets the requirements. As for the performance of the initial quality evaluation model, it can be determined by comparing the difference between the subjective image quality score carried by itself in step S101 and the objective image quality score output in S104. The smaller the difference between the two, it proves that the image output by the initial quality evaluation model The higher the consistency between the objective quality score and the subjective perception of the human eye, the higher the performance of the initial quality evaluation model.

In the technical solution provided by the embodiment of the present invention, the evaluation score of the subjective overall perception of the image by the human eye is used as the feature vector training model, and for the infrared image to be tested, an objective evaluation oriented to specific distortion with a high degree of consistency with the subjective perception of the human eye is selected. The method conducts objective evaluation, obtains the objective score value of the infrared image to be tested in each sub-dimension, inputs the objective score vector of each sub-dimension into the trained model, obtains the objective total score of the infrared image to be tested, and compares the objective score of the subjective score. Combined with the objective score, the final obtained objective infrared image quality score is highly consistent with the subjective perception of the human eye, which is very suitable for application scenarios where the current infrared imaging content is complex and diverse, and the original undistorted image cannot be obtained. After the model training is completed, in the entire image quality evaluation process, only the objective sub-dimension evaluation score of the infrared image to be evaluated needs to be calculated, and the infrared image quality evaluation result that is highly consistent with the subjective perception of the human eye can be obtained. The infrared image quality evaluation that is highly consistent with the subjective perception of the human eye has strong operability and better practicability.

In the above-mentioned embodiment, there is no limitation on how to perform step S101. In this embodiment, a method for generating training sets and test sets is provided, that is, the training of images carrying subjective scores of image quality is generated based on the original infrared image data set. The implementation of the set and test set can include the following steps:

A1: Obtain infrared images collected by multiple infrared thermal imaging devices in various types of application scenarios in different lighting environments to form the original infrared image dataset; the manufacturers, packaging methods and resolutions of each infrared thermal imaging device are different.

This step is to describe the selection and construction of infrared test scenarios. Since there is currently no complete quality evaluation database for infrared images, an infrared image database can be established in order to evaluate the quality of infrared images. On the basis of fully learning and understanding the principles and characteristics of infrared imaging, 25 groups of indoor scenes and 25 groups of outdoor scenes can be selected and constructed, including typical indoor scenes: electric lights, computers, monitors, water cups, power supplies, air conditioners, tables and chairs, green plants, People, etc., typical outdoor scenes: the sky, the ground, and natural scene objects of various temperatures such as pedestrians, vehicles, trees, and buildings. Considering that different detectors have a great influence on the imaging effect, in order to improve the robustness of different detectors, the infrared imaging equipment for infrared image data acquisition can choose a variety of non-contact devices produced by different manufacturers, with different packaging methods and different resolutions. For the refrigerated infrared thermal imaging device, the packaging method can be, for example, ceramic packaging or metal packaging. Through this step, multiple sets of indoor and outdoor scenes under different weather and different time periods can be constructed, and there are about 1000+ pieces of data covering different distortion conditions such as noise, motion blur, out-of-focus and non-uniformity.

A2: Obtain the subjective image quality score of each infrared image in the original infrared image dataset.

After building a typical infrared test scene in the previous step, and using infrared thermal imaging equipment to collect image data to construct an infrared image data set, the infrared image data set can be subjectively scored, including but not limited to uniformity, noise, clarity The expert sub-dimension scores of brightness, contrast, and dynamic range and the joint total score of experts and ordinary users for the entire image are used to obtain the subjective scoring results of the image.

A3: Based on the preset division ratio, each infrared image in the original infrared image dataset is divided into multiple training images and multiple test images.

A4: Generate a training set according to multiple training images and the subjective image quality scores corresponding to each training image.

A5: Generate a test set according to multiple test images and the subjective image quality scores corresponding to each test image.

The preset division ratio of the above steps can be flexibly selected according to actual needs, for example, the original infrared image data set can be divided into a training set and a test set according to 8:2. For example, the 1000 pieces of image data including various distortions of different degrees collected in the above steps can be divided into a training set and a test set according to a ratio of 8:2, and the image data of the training set and the image data of the test set are divided. There is no intersection.

The above embodiment does not limit the implementation of subjective scoring of infrared image data sets. The present application also provides an optional implementation, which may include the following content:

Obtaining the subjective evaluation scores of multiple experts on each subjective quality index of each infrared image in the original infrared image data set, so as to obtain the subjective sub-dimension evaluation score of each infrared image;

Obtain the subjective evaluation scores of multiple experts and multiple ordinary users on the overall infrared image in the original infrared image data set, and calculate the subjective overall evaluation score of each infrared image according to the preset expert weight coefficient and user weight coefficient;

The subjective image quality score of each infrared image in the original infrared image dataset is determined according to the subjective sub-dimension evaluation score and the subjective overall evaluation score of each infrared image.

对原始红外图像数据集的红外图像进行整体主观评测的过程可为：整体主观评测分数包括专家评分分数和普通用户评分分数，例如可选取24位专家和50位普通用户，在得到所有观测者的打分结果后，进行加权均值归一化处理，专家打分权重可为0.6，普通用户打分权重可为0.4，以此得到数据集中所有图像的主观总分数表{z₁,z₂,z₃...z_n}。最终原始红外图像数据集中各红外图像的主观子维度评分结果向量和主观总评分值也即A2的图像质量主观分数可表示为

In this embodiment, the subjective scores of each infrared image in the original infrared image data set include expert sub-dimension scores and a joint weighted total score of experts and ordinary users. Among them, the visual perception characteristics of the human eye and the characteristics of infrared imaging can be combined to select several items that affect the quality of infrared images, as follows: uniformity, noise, sharpness/blur, contrast, and dynamic brightness range as subjective scoring sub-dimensions. Correspondingly, the subjective sub-dimension evaluation score may include a subjective score of non-uniformity, a subjective score of image noise, a subjective score of image clarity, a subjective score of image contrast, and a subjective score of dynamic brightness range, that is, the subjective quality index of the above embodiment. As shown in Figure 2, for each training image of the training set, the subjective score sub-dimension of each training image may include non-uniformity subjective score, image noise subjective score, image sharpness subjective score, image contrast subjective score and dynamic brightness range subjective score. A sub-dimension is a subjective quality index. The subjective sub-dimension evaluation score can be a data set or a matrix. Each element is the subjective quality score corresponding to each subjective quality index. The subjective sub-dimension evaluation score can also be a score. The score is obtained by adding the subjective quality scores corresponding to each subjective quality index, or it can be obtained by weighting and summing the subjective quality scores corresponding to each subjective quality index. The weight factor of each objective quality index can be advanced according to the actual application scenario. It is assumed that each test image corresponds to a set of subjective sub-dimension evaluation scores. In the subjective sub-dimension scoring of each infrared image in the original infrared image data set, that is, to determine the subjective evaluation score of each subjective quality index, the single-stimulus continuous quality classification method can be used. The current image to be tested is then continuously scored according to the scoring table, and the quality evaluation of the image to be tested is obtained according to the score and the scoring time. The subjective evaluation scores of the subjective quality indicators are scored by experts, making full use of the experts' own professional knowledge in infrared image processing and evaluation to score the performance of each sub-dimension of the image or the performance of each subjective quality index in the data set. The method selected 24 One expert is used as a rater, and the scoring results of all raters are averaged to obtain the subjective score vector table of all images in each sub-dimension in the dataset, as shown in

The process of overall subjective evaluation of the infrared images of the original infrared image dataset can be as follows: the overall subjective evaluation score includes expert scoring scores and ordinary user scoring scores. For example, 24 experts and 50 ordinary users can be selected. After scoring the results, the weighted mean is normalized, the expert scoring weight can be 0.6, and the ordinary user scoring weight can be 0.4, so as to obtain the subjective total score table of all images in the dataset {z ₁ , z ₂ , z ₃ .. .z _n }. The subjective sub-dimension score result vector and subjective total score value of each infrared image in the final original infrared image dataset, that is, the image quality subjective score of A2 can be expressed as

It can be seen from the above that no public image data set is available for the current long-wave infrared image evaluation field in this embodiment. On the basis of fully learning and understanding the principles and characteristics of infrared imaging, this embodiment constructs an infrared image data set with subjective scores to fill in the There is no professional evaluation dataset available in the field of infrared imaging. In this embodiment, the subjective score of the infrared image data set is divided into sub-dimension score and total score. The sub-dimension score adopts the expert score, and the expert's own professional knowledge in infrared image processing and evaluation is fully utilized to evaluate the sub-dimension of the image in the data set. Performance is scored. The total score adopts the joint weighted score of experts and ordinary users, which not only uses the experience and knowledge of experts, but also effectively combines the public's perception of the image, so that the subjective score of infrared images is comprehensive, detailed, professional and reliable.

Based on the above embodiment, the present application also provides an optional implementation for S102, that is, the implementation process of obtaining the initial quality evaluation model by training the machine learning model with the training set may include:

Build the support vector regression model in advance, use the grid optimization algorithm encapsulated by the LIBSVM toolkit to calculate the hyperparameter γ, and obtain the hyperparameter ε and the hyperparameter C input by the user at the same time;

Construct the score feature vector according to the subjective sub-dimension evaluation score of each training image;

The score feature vector and the subjective overall evaluation score corresponding to each training image are input into the support vector regression model for training, and the initial quality evaluation model is obtained.

In this embodiment, the machine learning model adopts the SVR (Support Vector Regression, Support Vector Regression) model, and the SVR can use the kernel function to map the nonlinear inseparable features in low dimensions to linear separable features in high dimensions. Solve the few-shot training problem. This embodiment uses the ε-insensitive nonlinear regression ε-SVR to complete the mapping from the feature vector to the quality score. The specific implementation can refer to the LIBSVM package. The hyperparameters that need to be manually set are C and ε, and γ can be used in the package. The encapsulated grid optimization algorithm is obtained.

In the above-mentioned embodiment, there is no limitation on how to perform step S103. In this embodiment, an optional calculation method of the objective sub-dimension evaluation score is provided, as shown in FIG. 2, that is, by using the no-reference evaluation algorithm The implementation process of calculating the objective score of each objective quality index of each test image in the test set to determine the objective sub-dimension evaluation score of each test image may include the following steps:

B1: The non-uniformity objective score is obtained by calculating the image space variance of each test image and performing normalization processing.

In this embodiment, the non-uniformity of the infrared image may include fixed pattern noise, dark signal non-uniformity and light response non-uniformity; the fixed pattern noise means that the characteristic parameters of different pixels in a pixel array are different; The dark signal of different pixels is different, which is called the non-uniformity of dark signal; because the sensitivity of different pixels is different, it is called the non-uniformity of light response. In probability theory and statistics, variance is a measure of the degree of dispersion of a set of data, and variance can be used to describe all types of non-uniformity. The larger the variance, the more uneven the image; the smaller the variance, the better the uniformity of the image. Therefore, for the test set image, the result obtained by calculating the spatial variance of the image and normalizing it can be used as the non-uniformity objective score of the image. For an infrared image with M rows and N columns, the non-uniformity calculation formula can be:

Among them, ρ is a user-defined parameter value, which is used to represent the non-uniformity of the infrared image, and y _ij is the pixel value at the pixel (i, j).

B2: Determine the objective score of image noise by calculating the local normalized luminance coefficient of each test image.

It can be understood that the noise of the infrared image may include thermal noise, shot noise, 1/f noise, fixed pattern noise and fringe noise. Among them, thermal noise and shot noise can be regarded as white noise, 1/f noise intensity is inversely proportional to frequency, and it is fractal noise, fixed pattern noise and stripe noise belong to non-uniformity noise, and the non-uniformity factor has been dealt with in step B1 So this step is for thermal noise, shot noise, and 1/f noise. In the BRISQUE algorithm, the MSCN (Mean Subtracted Contrast Normalized, image local normalized luminance) coefficients have statistical characteristics sensitive to various degradations, and the visual quality that affects the image distortion can be predicted by quantifying the changes in the statistical characteristics.

Optionally, an optional calculation method for the objective score of image noise may include:

For each test image, calculate the local normalized luminance coefficient of the current test image, fit the local normalized luminance coefficient through the generalized Gaussian model, and obtain the fitting parameter mean and fitting parameter variance; calculate the local normalized luminance coefficient in The local normalized luminance coefficient neighborhood coefficients in multiple directions are obtained by fitting each local normalized luminance coefficient neighborhood coefficient by an asymmetric generalized Gaussian model; The multi-dimensional statistical features are extracted from the variance of the fitting parameters and each fitting parameter, and the multi-dimensional statistical features are input into the initial quality evaluation model to obtain the objective score of image noise of the current test image.

计算如下：In this embodiment, for a natural image, the MSCN coefficient histogram shows Gaussian distribution characteristics, and for a noisy image, the histogram is more stable. MSCN coefficient

The calculation is as follows:

In the formula, c is a constant, preventing the denominator from being 0, i∈1,2....M, j∈1,2....N represents the pixel spatial position, I(i,j) represents the intensity of the center pixel, μ(i,j) represents the mean value of the current local area, and σ(i,j) represents the variance of the current local area. ω={w _k,l |k=-K,....,K; l=-L,....,L} is a two-dimensional circularly symmetric Gaussian weighting function, K=L=3 represents the image segmentation block size.

The normalized luminance information calculated above can be fitted by using the existing Generalized Gaussian Distribution (GGD) to obtain the fitting parameter mean z and variance φ ² . In addition, in order to reflect the statistical characteristics of adjacent coefficients, on the basis of MSCN, the MSCN neighborhood coefficients in four directions of horizontal H, vertical V, main diagonal D ₁ , and sub-diagonal D ₂ are constructed:

考虑人类视觉具有多尺度性，在原尺度上和2倍下采样尺度上分别提取特征，可提取36个统计特征，之后通过初始质量评价模型建立从36个特征到图像质量分数的映射。The Asymmetric Generalized Gaussian Distribution (AGGD) in the prior art can be used to fit the neighborhood MSCN coefficients in the four directions to obtain the fitting parameters

Considering the multi-scale nature of human vision, the features are extracted on the original scale and the 2-fold downsampling scale respectively, and 36 statistical features can be extracted, and then the mapping from the 36 features to the image quality score is established through the initial quality evaluation model.

B3: Determine the objective score of image clarity by calculating the structural similarity between each test image and the corresponding reference image.

It can be understood that the sharpness of an image can be represented by the structural similarity between the target image x and the reference image y, and the structural similarity between the images can include a brightness comparison function, a contrast comparison function, and a structural information comparison function.

Among them, the brightness comparison function can be expressed as:

The contrast comparison function can be expressed as:

The structure information comparison function can be expressed as:

In the formula, C ₁ , C ₂ and C ₃ are constants, avoid the denominator to be 0, μ _x is the grayscale mean of image x, μy is the grayscale mean of image _y , _σx is the grayscale standard deviation of image x, σ _y is the gray standard deviation of the image y, and σ _xy is the covariance of the images x and y.

Based on the above content, the structural similarity of the images in this embodiment can be obtained by the relationship SSIM(x,y)=[l(x,y)] ^α ·[c(x,y)] ^β ·[s(x,y) )] ^γ is calculated, α, β, and γ control the weights of brightness, contrast, and structural information respectively, for example, α=β=γ=1. After the structural similarity of each infrared image is determined, the greater the structural similarity, the clearer the image to be tested is, and the objective score of image clarity of the image to be tested can be calculated based on the structural similarity.

B4: Calculate the image contrast objective score of each test image based on the image spatial frequency.

On the basis of analyzing the contrast sensitivity characteristics of the human visual system, the contrast sensitivity function can be defined as:

f_R，f_C分别为水平、垂直方向的空间频率。where f is the spatial frequency of the image,

f _R , f _C are the spatial frequencies in the horizontal and vertical directions, respectively.

In the formula, M is the number of rows of the image, and N is the number of columns of the image.

Since the spatial frequency of the image is directly related to the contrast sensitivity characteristics of the human visual system, the calculation method of the objective contrast score of the image can be determined based on the spatial frequency of the image and the actual application scene, and the objective image contrast score of the image to be tested can be obtained through this calculation method. .

B5: Calculate the objective score of the dynamic brightness range of each test image based on the maximum gray level of the image and the minimum gray level of the image.

The dynamic brightness range of an infrared thermal imager refers to the ability to adapt to the temperature changes of the scene in the shooting scene, and specifically refers to the range of changes in the brightness of the infrared image, that is, the adjustment range that represents the brightest and darkest in the image. For the test set images, the dynamic brightness range of the image can be obtained using the dynamic range calculation formula and normalized as the objective score of the dynamic brightness range. The dynamic range calculation formula can be expressed as:

In the formula, L _max is the maximum gray level of the image obtained by statistics, and L _min is the minimum gray level of the image.

B6: For each test image, normalize the objective score of non-uniformity, objective score of image noise, objective score of image clarity, objective score of image contrast and objective score of dynamic brightness range of the current test image to obtain the current test image The objective sub-dimension evaluation score of .

It can be seen from the above that in this embodiment, based on the subjective overall perception of the image by the human eye, it is a comprehensive result of perception of various dimensions, and the total score of the infrared image quality by the human eye is regarded as the mapping function of the scores of each sub-dimension, where each sub-dimension The contribution of the score to the overall total score is different, that is, the sub-dimension evaluation score score_all (including the objective sub-dimension evaluation score and the subjective sub-dimension evaluation score) = function(score_nuc (including the non-uniform objective score and non-uniform subjective score), score_noise (including objective score of image noise and subjective score of image noise), score_blur (including objective score of image sharpness and subjective score of image sharpness), score_contrast (objective score of image contrast and subjective score of image contrast), score_dynamic (including objective score of dynamic brightness range) score and dynamic brightness range subjective score)). For the infrared image to be tested, an objective evaluation method oriented to specific distortions that is highly consistent with the subjective perception of the human eye is selected for objective evaluation, and the objective score value of the infrared image to be tested in each sub-dimension is obtained, and the objective score vector of each sub-dimension is input. In the trained initial quality evaluation model, the objective total score of the infrared image to be tested is obtained, which realizes a powerful combination of the objectivity of the subjective score and the high efficiency of the objective score.

The above embodiment does not make any limitation on how to perform the process of determining the objective score of image clarity by calculating the structural similarity between each test image and the corresponding reference image. This embodiment also provides an optional implementation. Can include:

For each test image, use a Gaussian smoothing filter to perform low-pass filtering on the current test image to obtain a corresponding current reference image;

Extract the gradient information of the current test image and the current reference image and the edge information in the target direction respectively;

Generate a test gradient image according to the gradient information and edge information of the current test image; generate a reference gradient image according to the gradient information and edge information of the current reference image;

Determine a plurality of target image blocks that satisfy the preset gradient information conditions from the test gradient image, and determine that each target image block corresponds to the target reference image block in the reference gradient image;

Call the pre-built image structure similarity relationship formula to calculate the structural similarity between each target image block and the corresponding target reference image block; the image structure similarity relationship formula is based on the brightness comparison function, the contrast comparison function and the structure information comparison function and their respective The weight coefficient is determined;

The objective score of the image clarity of the current test image is determined according to the structural similarity of each target image block.

In this embodiment, the non-reference image sharpness evaluation index NRSS is introduced to calculate the image sharpness. First, a reference image is constructed for the image to be evaluated, and the image to be evaluated is defined as I. For example, a Gaussian smoothing filter with a size of 7×7 and σ ² =6 can be used to low-pass filter the image I to be evaluated to obtain the reference image of the image I to be evaluated. Image I _r =LPF(I). Extract the gradient information of the images I and I _r , using the feature that the human eye is most sensitive to the edge information in the horizontal and vertical directions, the Sobel operator can be used to extract the edge information in the horizontal and vertical directions respectively, and generate gradients based on the edge information and gradient information. image, the gradient images defining I and I _r are G and G _r , respectively.

The process of determining multiple target image blocks that meet the preset gradient information conditions may be: the preset gradient information in this embodiment may be the most abundant gradient information. Specifically, a gradient information threshold may be determined first, and the gradient information greater than the threshold The image patch is the gradient patch with the richest gradient information. As an optional implementation manner, the process of selecting N image blocks with the most abundant gradient information from the gradient image G may be: dividing the gradient image G into multiple small image blocks according to 8*8, in order to avoid losing important For the edge, the inter-block step size of each image block is 4, that is, the adjacent blocks have 50% overlap. Calculate the variance of each image block. The larger the variance is, the richer the gradient information is. Find the N blocks with the largest variance, denoted as {x _i |i=1,2,...,N}, in the corresponding G _r The corresponding block of is defined as {y _i |i=1,2,...,N}. Of course, other methods can also be used to select the desired target image block from the gradient image, which does not affect the implementation of the present application. After the target image block is determined, the structural similarity SSIM(x _i , y _i ) SSIM(x _i , y _i ) of each target image block _xi and its corresponding image block yi in the reference image can be calculated first _. The image structure similarity relationship SSIM(x,y)=[l(x,y)] ^α ·[c(x,y)] ^β ·[s(x,y)] ^γ is calculated, and the structure is obtained after calculation After similarity, the objective score of image clarity can be calculated based on the following structural clarity NRSS calculation relationship, and the structural clarity NRSS calculation relationship can be expressed as:

It can be seen from the above that in the process of calculating the objective score of image clarity in this embodiment, the structural similarity is calculated by selecting several representative image blocks from the test image and the reference image, which can not only improve the overall image quality evaluation efficiency, but also improve the overall image quality evaluation efficiency. The calculation accuracy of the objective score of image clarity is improved, which is beneficial to improve the accuracy of image quality evaluation.

The above-mentioned embodiment does not make any limitation on how to calculate the objective image contrast score of each test image based on the image spatial frequency. The present embodiment also provides an optional calculation method for the image contrast objective score, which may include the following content:

For each test image, use the preset size template to traverse the current test image according to the spatial frequency in the horizontal direction and the vertical direction, and obtain the spatial frequency matrix of the current test image;

Based on the spatial frequency matrix, the image spatial frequency of each pixel is normalized one by one, and the normalized image spatial frequency matrix is obtained;

Determine the contrast sensitivity weight matrix of the current test image according to the pre-built contrast sensitivity relationship and the normalized image spatial frequency matrix;

The objective score of the image contrast of the current test image is calculated according to the contrast sensitivity weight matrix, the image size of the current test image, the maximum gray level and the minimum gray level.

In this embodiment, a 3*3 template can be used to traverse the image according to the spatial frequency f _R in the horizontal direction and the spatial frequency f _C in the vertical direction to obtain the image spatial frequency matrix f(i, j), and based on the normalization The calculation relationship normalizes the image spatial frequency one by one, and the normalized calculation relationship can be expressed as:

In the formula, f _min and f _max are the minimum and maximum values of the image spatial frequency f(i, j), after obtaining the normalized image spatial frequency matrix f _mon (i, j), f _mon (i, j) can be ,j) into the contrast sensitivity function to obtain the image contrast sensitivity weight matrix C(i,j). Based on the contrast sensitivity weight matrix, the objective image contrast score of the test image is obtained by calling the pre-built contrast objective score calculation relation. The calculation relationship of the contrast objective score can be expressed as:

L_max为统计得到的画面的最大灰度级，L_min为图像的最小灰度级，最大灰度级和最小灰度级分别为统计的图像中的最大像素值和最小像素值所得。M为当前测试图像的行数，N为当前测试图像的列数，m为第m行，n为第n列。in,

L _max is the maximum gray level of the image obtained by statistics, L _min is the minimum gray level of the image, and the maximum gray level and the minimum gray level are obtained from the maximum pixel value and the minimum pixel value in the statistical image, respectively. M is the number of rows of the current test image, N is the number of columns of the current test image, m is the mth row, and n is the nth column.

It can be seen from the above that this embodiment can improve the accuracy of calculating the objective contrast score of the image by reflecting the spatial frequency of the sensitivity of the human eye and combining the image grayscale information to calculate the objective contrast score, thereby improving the accuracy of image quality evaluation.

The above-mentioned embodiment does not make any limitation on how to perform S105. The present application also provides a method for determining the final image quality evaluation model based on the initial quality evaluation model, according to the image quality objective score of each test image and the corresponding image quality subjective score. Implementation methods can include:

For each test image, the specific values of the performance measurement indicators of the objective image quality score and the subjective image quality score of the current test image are calculated respectively. The performance measurement indicators are Pearson's linear correlation coefficient, Spearman's rank correlation coefficient, and Kendall's rank correlation coefficient. Any one or any combination of coefficient and root mean square error, correspondingly, is a performance measure for calculating the objective image quality score and the subjective image quality score of the current test image respectively.

If the Pearson linear correlation coefficient and/or the Spearman rank correlation coefficient and/or the Kendall rank correlation coefficient and/or the root mean square error meet the preset performance conditions, that is, the performance measure meets the preset performance conditions, the initial The quality evaluation model is used as an image quality evaluation model;

If the Pearson linear correlation coefficient and/or the Spearman rank correlation coefficient and/or the Kendall rank correlation coefficient and/or the root mean square error do not meet the preset performance conditions, that is, the performance measure does not meet the preset performance conditions, then Generate an instruction to optimize the initial quality evaluation model, and retrain the initial quality evaluation model until the preset performance conditions are met.

Among them, the Pearson linear correlation coefficient PLCC of the objective image quality score and the subjective image quality score can be calculated by the Pearson linear correlation coefficient calculation relationship. high, indicating that the performance of the initial quality evaluation model is better. Pearson's linear correlation coefficient calculation relationship can be expressed as:

和

分别是{x₁,x₂,…,x_n}和{y₁,y₂,…y_n}的均值，σ_x和σ_y分别为其标准差。Among them, x _i ,i∈{1,2,....n} represents the image quality objective score array for the test set images, y _i ,i∈{1,2,....n} represents the test set The image quality subjective score array of the image, n is the number of images in the test set,

and

are the mean values of {x ₁ , x ₂ ,…,x _n } and {y ₁ ,y ₂ ,… y _n }, respectively, and σ _x and σ _y are their standard deviations, respectively.

In this implementation, the Spearman rank correlation coefficient SROCC of the objective image quality score and the subjective image quality score can be calculated by the Spearman rank correlation coefficient calculation relationship. The larger the value of the Spearman rank correlation coefficient, the model output result The higher the correlation with the human eye score, the better the performance of the initial quality evaluation model. The formula for calculating the Spearman rank correlation coefficient can be expressed as:

Among them, n is the number of images in the test set, sort the array x _i and the array y _i according to the values in the group from small to large, respectively, and get r _xi and ry _yi are the order of the i-th value in the array, r _xi -ry _yi The order difference of the two groups of data.

In this implementation, the Kendall rank correlation coefficient KROCC of the objective image quality score and the subjective image quality score may be calculated through a Kendall rank correlation coefficient calculation relation. The larger the Kendall rank correlation coefficient, the higher the correlation between the model output and the human eye score, indicating that the performance of the initial quality evaluation model is better. For the objective total score array x _i ,i∈{1,2,....n} for the test set images, the human eye subjective score array y _i ,i∈{1,2,... .n}, define that the data pairs in the two arrays are consistent (x _i >x _j , y _i >y _j or x _i <x _j , y _i <y _j ) There are P data pairs, the data pairs in the two arrays There are Q data pairs that are inconsistent (x _i >x _j , y _i <y _j or x _i <x _j , y _i >y _j ), and the Kendall rank correlation coefficient calculation relation can be expressed as:

where n is the number of images in the test set.

In this implementation, the root mean square error of the objective image quality score and the subjective image quality score can be calculated by using the root mean square error calculation relationship. The smaller the root mean square error, the higher the correlation between the model output and the human eye score, indicating that the performance of the initial quality evaluation model is better. The root mean square error calculation relationship can be expressed as:

Among them, x _i ,i∈{1,2,....n} represents the objective total score array for the test set images, y _i ,i∈{1,2,....n} represents the test set images is an array of subjective human eye ratings, and n is the number of images in the test set.

For example, Table 1 shows the numerical values of Pearson's linear correlation coefficient, Spearman's rank correlation coefficient, Kendall's rank correlation coefficient and root mean square error in an illustrative example.

Table 1 Specific values of performance metrics

index PLCC SROCC KROCC RMSE consistency 0.9302 0.9123 0.9541 2.6517

In this embodiment, the preset performance condition refers to the accuracy of the quality evaluation model determined by those skilled in the art according to the actual application scenario. The combination of the Pearman rank correlation coefficient, the Kendall rank correlation coefficient and the root mean square error is related. If the performance measurement index is the root mean square error, the preset performance condition can be that the root mean square error value is less than 2.7. If calculated according to the above method If the obtained root mean square error value is greater than 2.7, the performance of the initial quality evaluation model does not meet the standard and needs to be optimized. If the root mean square error value calculated in the above method is less than 2.7, the initial quality evaluation model meets the preset performance conditions. If the performance measurement indicators are the root mean square error and the Pearson linear correlation coefficient, the preset performance conditions can be that the root mean square error value is less than 2.7 and the Pearson linear correlation coefficient is greater than 0.9. If the value is greater than 2.7 and/or the Pearson linear correlation coefficient is less than 0.9, the performance of the initial quality evaluation model is not up to standard and needs to be optimized. If the root mean square error value calculated in the above method is less than 2.7 and the Pearson linear correlation coefficient is greater than 0.9, Then the initial quality evaluation model satisfies the preset performance conditions.

It can be seen from the above that in this embodiment, by comparing the subjective total score of the test set image and the objective total score of the Spearman order correlation coefficient SROCC, Pearson linear correlation coefficient PLCC, Kendall order correlation coefficient KROCC, and root mean square error RMSE, the objective of the infrared image quality is achieved. The evaluation is consistent with the subjective perception of the human eye, which is highly practical.

It can be understood that the above-mentioned embodiment describes how to obtain a model for actual infrared image quality evaluation, so that it outputs an image quality evaluation result that is highly consistent with the subjective perception of the human eye, so as to realize the image quality evaluation of infrared images. Evaluation, based on this, this application also provides another embodiment. This embodiment is used to implement image quality evaluation on any infrared image with an unknown subjective image quality score. Please refer to FIG. 3, which may include the following content:

S301: Obtain an image quality evaluation model in advance by using the steps described in any one of the above infrared image quality evaluation method embodiments.

The image quality evaluation model in this step is the final image quality evaluation model obtained in S105 of the above embodiment.

S302: Acquire an infrared image to be evaluated.

The infrared image to be evaluated is any infrared image that needs to be evaluated for image quality.

S303: Determine the objective sub-dimension evaluation score of the infrared image to be evaluated by calculating the objective score of each objective quality index of the infrared image to be evaluated by using the no-reference evaluation algorithm.

S304: Input the objective sub-dimension evaluation score into the image quality evaluation model to obtain the image quality evaluation score of the infrared image to be evaluated.

For the methods or steps in this embodiment that are the same as the above-mentioned embodiments, reference may be made to the content recorded in the above-mentioned embodiments, which will not be repeated here.

It can be seen from the above that this embodiment constructs a reference-free infrared image quality evaluation method for non-specific distortion types, through which the infrared image quality evaluation results of infrared images under different infrared detection devices and different scenarios can be obtained, and the results are obtained. It is highly consistent with the subjective perception of the human eye.

It should be noted that there is no strict sequence of execution between the steps in this application. As long as the logical sequence is followed, these steps can be executed at the same time or in a preset sequence. This is a schematic way, and does not mean that there can only be such an execution order.

The embodiment of the present invention also provides a corresponding device for the infrared image quality evaluation method, which further makes the method more practical. Wherein, the device can be described from the perspective of functional modules and the perspective of hardware. The infrared image quality evaluation device provided by the embodiment of the present invention is introduced below. The infrared image quality evaluation device described below and the infrared image quality evaluation method described above can be referred to each other correspondingly.

From the perspective of functional modules, see FIG. 4 . FIG. 4 is a structural diagram of an infrared image quality evaluation device provided by an embodiment of the present invention in a specific implementation manner. The device may include:

A data set construction module 401, used for generating a training set and a test set of subjective scores of image-carrying image quality based on the original infrared image data set;

The model training module 402 is used to train the machine learning model by using each training image in the training set and the corresponding subjective image quality score to obtain an initial quality evaluation model;

The sub-dimension score calculation module 403 is used to calculate the objective score of each objective quality index of each test image in the test set by using a no-reference evaluation algorithm, so as to determine the objective sub-dimension evaluation score of each test image;

The objective evaluation module 404 inputs the evaluation scores of each objective sub-dimension into the initial quality evaluation model to obtain the image quality objective score of each test image;

The model determination module 405 is configured to determine the final image quality evaluation model according to the objective image quality scores and corresponding subjective image quality scores of each test image based on the initial quality evaluation model.

Optionally, in some implementations of this embodiment, the above-mentioned data set construction module 401 may be used to: acquire infrared images collected by multiple infrared thermal imaging devices in various types of application scenarios under different lighting environments, so as to form the original image. Infrared image data set; the manufacturers, packaging methods and resolutions of each infrared thermal imaging equipment are different; the subjective score of the image quality of each infrared image in the original infrared image data set is obtained; The images are divided into multiple training images and multiple test images; the training set is generated according to the multiple training images and the subjective image quality scores corresponding to each training image; the test set is generated according to the multiple test images and the image quality subjective scores corresponding to each test image .

As an optional implementation of the above-mentioned embodiment, the above-mentioned data set construction module 401 may be further configured to: obtain the subjective evaluation scores of multiple experts for each subjective quality index of each infrared image, so as to obtain the subjective evaluation score of each infrared image. Subjective sub-dimension evaluation score; obtain the overall subjective evaluation scores of each infrared image by multiple experts and multiple ordinary users, and calculate the subjective overall evaluation score of each infrared image according to the preset expert weight coefficient and user weight coefficient; according to each infrared image The subjective sub-dimension evaluation score and the subjective overall evaluation score of , determine the image quality subjective score of each infrared image.

As another optional implementation of the above-mentioned embodiment, the above-mentioned model training module 402 may be further used for: constructing a support vector regression model in advance; constructing a score feature vector according to the subjective sub-dimension evaluation scores of each training image; The value feature vector and the subjective overall evaluation score corresponding to each training image are input to the support vector regression model for training, and the initial quality evaluation model is obtained.

Optionally, in other implementations of this embodiment, the sub-dimension score calculation module 403 may include:

The uniformity calculation unit is used to obtain the non-uniformity objective score by calculating the image space variance of each test image and performing normalization processing;

a noise calculation unit, configured to determine the objective score of image noise by calculating the local normalized luminance coefficient of each test image;

A sharpness calculation unit, used for determining the objective image sharpness score by calculating the structural similarity between each test image and the corresponding reference image;

a contrast calculation unit, used for calculating the objective image contrast score of each test image based on the image spatial frequency;

a dynamic range calculation unit, used for calculating the objective score of the dynamic brightness range of each test image based on the maximum gray level of the image and the minimum gray level of the image;

The sub-dimension overall score calculation unit is used to normalize the non-uniformity objective score, image noise objective score, image clarity objective score, image contrast objective score and dynamic brightness range objective score of the current test image for each test image. process to obtain the objective sub-dimension evaluation score of the current test image.

As an optional implementation of the above-mentioned embodiment, the above-mentioned noise calculation unit may be further configured to: for each test image, calculate the local normalized luminance coefficient of the current test image, and fit the local normalized brightness coefficient by a generalized Gaussian model Luminance coefficient, get the fitting parameter mean and fitting parameter variance; calculate the local normalized luminance coefficient neighborhood coefficients of local normalized luminance coefficients in multiple directions, and fit each local normalized by an asymmetric generalized Gaussian model The brightness coefficient and the neighborhood coefficient are used to obtain multiple fitting parameters; the multi-dimensional statistical features are extracted from the fitting parameter mean, fitting parameter variance and each fitting parameter at different scales, and the multi-dimensional statistical features are input into the initial quality evaluation model to obtain The objective score of image noise for the current test image.

As another optional implementation of the above-mentioned embodiment, the above-mentioned sharpness calculation unit may be further configured to: for each test image, use a Gaussian smoothing filter to perform low-pass filtering on the current test image to obtain a corresponding current reference image ; Extract the gradient information of the current test image and the current reference image and the edge information in the target direction respectively; Generate a test gradient image according to the gradient information and edge information of the current test image; Generate a reference gradient image according to the gradient information and edge information of the current reference image ; Determine multiple target image blocks that meet the preset gradient information conditions from the test gradient image, and determine that each target image block corresponds to the target reference image block in the reference gradient image; call the pre-built image structure similarity relationship to calculate each target The structural similarity between the image block and the corresponding target reference image block; the image structure similarity relationship is determined according to the brightness comparison function, the contrast comparison function and the structural information comparison function and their respective weight coefficients; according to the structural similarity of each target image block Determines the objective score of image clarity for the current test image.

As another optional implementation of the above-mentioned embodiment, the above-mentioned contrast calculation unit may be further configured to: for each test image, use a preset size template to traverse the current test image according to the spatial frequency in the horizontal direction and the vertical direction, Obtain the spatial frequency matrix of the current test image; based on the spatial frequency matrix, normalize the image spatial frequency of each pixel point one by one to obtain the normalized image spatial frequency matrix; and the normalized image space frequency matrix to determine the contrast sensitivity weight matrix of the current test image; calculate the current test image according to the contrast sensitivity weight matrix, the image size of the current test image, the maximum gray level and the minimum gray level objective score of image contrast.

Optionally, in some other implementations of this embodiment, the above-mentioned model determination module 405 may be further configured to: for each test image, respectively calculate the performance measurement of the objective image quality score and the subjective image quality score of the current test image. indicators; wherein, the performance measurement indicators include one or more of Pearson's linear correlation coefficient, Spearman's rank correlation coefficient, Kendall's rank correlation coefficient, and root mean square error; if the performance measurement indicators meet the preset performance conditions, then The initial quality evaluation model is used as the image quality evaluation model; if the performance measurement index does not meet the preset performance conditions, an instruction to optimize the initial quality evaluation model is generated, and the initial quality evaluation model is retrained until the preset performance conditions are met.

From the perspective of functional modules, please refer to FIG. 5 . FIG. 5 is a structural diagram of an infrared image quality evaluation apparatus provided in an embodiment of the present invention in another specific implementation manner. The apparatus may include:

A model building module 501, configured to obtain an image quality evaluation model by using any one of the above infrared image quality evaluation methods in advance;

an image acquisition module 502, configured to acquire an infrared image to be evaluated;

The objective scoring module 503 is configured to calculate the objective score of each objective quality index of the infrared image to be evaluated by using the no-reference evaluation algorithm, so as to determine the objective sub-dimension evaluation score of the infrared image to be evaluated;

The quality evaluation module 504 is configured to input the objective sub-dimension evaluation scores into the image quality evaluation model to obtain the image quality evaluation scores of the infrared images to be evaluated.

The functions of each functional module of the infrared image quality evaluation apparatus according to the embodiment of the present invention may be specifically implemented according to the methods in the foregoing method embodiments, and the specific implementation process may refer to the relevant descriptions of the foregoing method embodiments, which will not be repeated here.

It can be seen from the above that the embodiments of the present invention can efficiently and conveniently implement infrared image quality evaluation that is highly consistent with the subjective perception of the human eye.

The infrared image quality evaluation device mentioned above is described from the perspective of functional modules. Further, the present application also provides an electronic device, which is described from the perspective of hardware. FIG. 6 is a schematic structural diagram of an electronic device provided in an embodiment of the present application in an implementation manner. As shown in FIG. 6 , the electronic device includes a memory 60 for storing a computer program; a processor 61 for implementing the steps of the infrared image quality evaluation method mentioned in any of the above embodiments when executing the computer program.

The processor 61 may include one or more processing cores, such as a 4-core processor or an 8-core processor, and the processor 61 may also be a controller, a microcontroller, a microprocessor, or other data processing chips. The processor 61 can be implemented by at least one hardware form among DSP (Digital Signal Processing, digital signal processing), FPGA (Field-Programmable Gate Array, field programmable gate array), and PLA (Programmable Logic Array, programmable logic array). . The processor 61 may also include a main processor and a coprocessor. The main processor is a processor used to process data in the wake-up state, also called a CPU (Central Processing Unit, central processing unit); the coprocessor is a A low-power processor for processing data in a standby state. In some embodiments, the processor 61 may be integrated with a GPU (Graphics Processing Unit, image processor), and the GPU is used for rendering and drawing the content that needs to be displayed on the display screen. In some embodiments, the processor 61 may further include an AI (Artificial Intelligence, artificial intelligence) processor, where the AI processor is used to process computing operations related to machine learning.

Memory 60 may include one or more computer-readable storage media, which may be non-transitory. Memory 60 may also include high-speed random access memory as well as non-volatile memory, such as one or more magnetic disk storage devices, flash storage devices. The memory 60 may in some embodiments be an internal storage unit of an electronic device, such as a hard disk of a server. In other embodiments, the memory 60 may also be an external storage device of the electronic device, such as a plug-in hard disk equipped on a server, a Smart Media Card (SMC), a Secure Digital (SD) card, and a flash memory card. (Flash Card) etc. Further, the memory 60 may also include both an internal storage unit of the electronic device and an external storage device. The memory 60 can not only be used to store application software installed in the electronic device and various types of data, such as code of a program executing the vulnerability processing method, etc., but also can be used to temporarily store data that has been output or will be output. In this embodiment, the memory 60 is at least used to store the following computer program 601, wherein, after the computer program is loaded and executed by the processor 61, the relevant steps of the infrared image quality evaluation method disclosed in any of the foregoing embodiments can be implemented. In addition, the resources stored in the memory 60 may also include an operating system 602, data 603, etc., and the storage mode may be short-term storage or permanent storage. The operating system 602 may include Windows, Unix, Linux, and the like. The data 603 may include, but is not limited to, data corresponding to the infrared image quality evaluation result, and the like.

In some embodiments, the above electronic device may further include a display screen 62 , an input/output interface 63 , a communication interface 64 or a network interface, a power supply 65 and a communication bus 66 . Among them, the display screen 62 and the input and output interface 63 such as a keyboard (Keyboard) belong to the user interface, and the optional user interface may also include a standard wired interface, a wireless interface, and the like. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode, organic light-emitting diode) touch device, and the like. The display, also appropriately referred to as a display screen or display unit, is used to display information processed in the electronic device and to display a visual user interface. The communication interface 64 may optionally include a wired interface and/or a wireless interface, such as a WI-FI interface, a Bluetooth interface, etc., and is generally used to establish a communication connection between an electronic device and other electronic devices. The communication bus 66 may be a peripheral component interconnect (PCI for short) bus or an extended industry standard architecture (extended industry standard architecture, EISA for short) bus or the like. The bus can be divided into address bus, data bus, control bus and so on. For ease of presentation, only one thick line is used in FIG. 6, but it does not mean that there is only one bus or one type of bus.

Those skilled in the art can understand that the structure shown in FIG. 6 does not constitute a limitation on the electronic device, and may include more or less components than the one shown, for example, may also include sensors 67 that implement various functions.

The functions of each functional module of the electronic device in this embodiment of the present invention may be specifically implemented according to the methods in the foregoing method embodiments, and the specific implementation process may refer to the relevant descriptions of the foregoing method embodiments, which will not be repeated here.

It can be seen from the above that the embodiments of the present invention can efficiently and conveniently implement infrared image quality evaluation that is highly consistent with the subjective perception of the human eye.

It can be understood that, if the infrared image quality evaluation method in the above embodiment is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the present application can be embodied in the form of software products in essence, or the parts that contribute to the prior art, or all or part of the technical solutions, and the computer software products are stored in a storage medium , to execute all or part of the steps of the methods in the various embodiments of the present application. The aforementioned storage media include: U disk, mobile hard disk, read-only memory (Read-Only Memory, ROM), random access memory (Random Access Memory, RAM), electrically erasable programmable ROM, registers, hard disks, multimedia Cards, card-type memories (such as SD or DX memories, etc.), magnetic memories, removable disks, CD-ROMs, magnetic disks, or optical disks, and other media that can store program codes.

Based on this, an embodiment of the present invention further provides a readable storage medium storing a computer program, and when the computer program is executed by a processor, the steps of the infrared image quality evaluation method described in any one of the above embodiments are performed.

The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments, and the same or similar parts between the various embodiments may be referred to each other. As for the hardware disclosed in the embodiments, including the apparatus and electronic equipment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant part can be referred to the description of the method.

Professionals may further realize that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two, in order to clearly illustrate the possibilities of hardware and software. Interchangeability, the above description has generally described the components and steps of each example in terms of functionality. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Skilled artisans may implement the described functionality using different methods for each particular application, but such implementations should not be considered beyond the scope of the present invention.

The infrared image quality evaluation method, device, electronic device and readable storage medium provided by the present application have been described in detail above. The principles and implementations of the present invention are described herein by using specific examples, and the descriptions of the above embodiments are only used to help understand the method and the core idea of the present invention. It should be pointed out that for those skilled in the art, without departing from the principle of the present invention, several improvements and modifications can also be made to the present application, and these improvements and modifications also fall within the protection scope of the claims of the present application.

Claims (14)

1. an infrared image quality evaluation method, is characterized in that, comprises: Generate training sets and test sets with images carrying subjective scores of image quality based on the original infrared image dataset; By using each training image in the training set and the corresponding subjective image quality score to train the machine learning model, an initial quality evaluation model is obtained; Determine the objective sub-dimension evaluation score of each test image by calculating the objective score of each objective quality index of each test image in the test set by using a no-reference evaluation algorithm; Inputting each objective sub-dimension evaluation score into the initial quality evaluation model to obtain the image quality objective score of each test image; Based on the initial quality evaluation model, the final image quality evaluation model is determined according to the objective image quality scores of each test image and the corresponding subjective image quality scores. 2. infrared image quality evaluation method according to claim 1, is characterized in that, described based on original infrared image data set, the training set and the test set that the image carries the subjective score of image quality, comprises: Obtain infrared images collected by multiple infrared thermal imaging devices in various types of application scenarios under different lighting environments to form the original infrared image data set; wherein, the manufacturer, packaging method and resolution of each infrared thermal imaging device different; obtaining the subjective image quality score of each infrared image in the original infrared image dataset; Dividing each infrared image in the original infrared image data set into multiple training images and multiple testing images based on a preset division ratio; Generate a training set according to multiple training images and the subjective image quality scores corresponding to each training image; A test set is generated according to multiple test images and the subjective image quality scores corresponding to each test image. 3. The infrared image quality evaluation method according to claim 2, wherein the acquiring the subjective image quality score of each infrared image in the original infrared image data set comprises: Obtain the subjective evaluation scores of each subjective quality index of each infrared image by multiple experts, so as to obtain the subjective sub-dimension evaluation score of each infrared image; Obtain the overall subjective evaluation scores of each infrared image by multiple experts and multiple ordinary users, and calculate the subjective overall evaluation score of each infrared image according to the preset expert weight coefficient and user weight coefficient; The subjective image quality score of each infrared image is determined according to the subjective sub-dimension evaluation score and the subjective overall evaluation score of each infrared image. 4. The infrared image quality evaluation method according to claim 3, wherein the training machine learning model by utilizing each training image in the training set and the corresponding image quality subjective score to obtain an initial quality evaluation model, comprising: Pre-built support vector regression models; Construct the score feature vector according to the subjective sub-dimension evaluation score of each training image; The score feature vector and the subjective overall evaluation score corresponding to each training image are input into the support vector regression model for training to obtain the initial quality evaluation model. 5. The infrared image quality evaluation method according to any one of claims 1 to 4, wherein the objective score of each objective quality index of each test image in the test set is calculated by using a reference-free evaluation algorithm, Determine objective sub-dimension evaluation scores for each test image, including: The non-uniformity objective score is obtained by calculating the image space variance of each test image and normalizing it; Determine the objective score of image noise by calculating the local normalized luminance coefficient of each test image; Determine the objective score of image clarity by calculating the structural similarity between each test image and the corresponding reference image; Calculate the image contrast objective score of each test image based on the image spatial frequency; Calculate the objective score of the dynamic brightness range of each test image based on the maximum gray level of the image and the minimum gray level of the image; For each test image, normalize the non-uniformity objective score, image noise objective score, image clarity objective score, image contrast objective score, and dynamic brightness range objective score of the current test image to obtain the current test image The objective sub-dimension evaluation score of . 6. The infrared image quality evaluation method according to claim 5, characterized in that, determining the objective score of image noise by calculating the local normalized luminance coefficient of each test image, comprising: For each test image, calculate the local normalized luminance coefficient of the current test image, and fit the local normalized luminance coefficient through a generalized Gaussian model to obtain the fitting parameter mean and fitting parameter variance; calculating the local normalized luminance coefficient neighborhood coefficients of the local normalized luminance coefficient in multiple directions, and fitting each local normalized luminance coefficient neighborhood coefficient by an asymmetric generalized Gaussian model to obtain a plurality of fitting parameters; Extract multi-dimensional statistical features from the fitting parameter mean, the fitting parameter variance, and each fitting parameter at different scales, and input the multi-dimensional statistical features into the initial quality evaluation model to obtain the current test image. Image noise objective score. 7. The infrared image quality evaluation method according to claim 5, wherein, determining the objective image clarity score by calculating the structural similarity between each test image and the corresponding reference image, comprising: For each test image, a Gaussian smoothing filter is used to perform low-pass filtering on the current test image to obtain the corresponding current reference image; respectively extracting the gradient information of the current test image and the current reference image and the edge information in the target direction; Generate a test gradient image according to the gradient information and edge information of the current test image; generate a reference gradient image according to the gradient information and edge information of the current reference image; Determine from the test gradient image a plurality of target image blocks that satisfy the preset gradient information conditions, and determine that each target image block corresponds to a target reference image block in the reference gradient image; Calling the pre-built image structure similarity relationship formula to calculate the structure similarity between each target image block and the corresponding target reference image block; wherein, the image structure similarity relationship formula is based on the brightness comparison function, the contrast comparison function and the structure information. The comparison functions and their respective weight coefficients are determined; The objective image definition score of the current test image is determined according to the structural similarity of each target image block. 8. The infrared image quality evaluation method according to claim 5, characterized in that, calculating the objective image contrast score of each test image based on the image spatial frequency, comprising: For each test image, use a preset size template to traverse the current test image according to the spatial frequency of the horizontal direction and the vertical direction, and obtain the spatial frequency matrix of the current test image; Based on the spatial frequency matrix, normalize the image spatial frequency of each pixel point one by one to obtain a normalized image spatial frequency matrix; Determine the contrast sensitivity weight matrix of the current test image according to the pre-built contrast sensitivity relationship and the normalized image spatial frequency matrix; The objective image contrast score of the current test image is calculated according to the contrast sensitivity weight matrix, the image size, the maximum gray level and the minimum gray level of the current test image. 9. The infrared image quality evaluation method according to any one of claims 1 to 4, wherein, based on the initial quality evaluation model, according to the objective image quality score of each test image and the corresponding subjective image quality score Determine the final image quality evaluation model, including: For each test image, the performance measurement indicators of the objective image quality score and the subjective image quality score of the current test image are respectively calculated; wherein, the performance measurement indicators include Pearson linear correlation coefficient, Spearman rank correlation coefficient, Kendall rank One or more of correlation coefficient, root mean square error; If the performance measurement index satisfies a preset performance condition, the initial quality evaluation model is used as the image quality evaluation model; If the performance measurement index does not meet the preset performance condition, an instruction to optimize the initial quality evaluation model is generated, and the initial quality evaluation model is retrained until the preset performance condition is met. 10. A method for evaluating infrared image quality, comprising: Using the infrared image quality evaluation method according to any one of claims 1 to 9 to obtain an image quality evaluation model in advance; Obtain the infrared image to be evaluated; Determine the objective sub-dimension evaluation score of the infrared image to be evaluated by calculating the objective score of each objective quality index of the infrared image to be evaluated by using the no-reference evaluation algorithm; The objective sub-dimension evaluation score is input into the image quality evaluation model to obtain the image quality evaluation score of the infrared image to be evaluated. 11. A device for evaluating infrared image quality, comprising: A dataset construction module, which is used to generate a training set and a test set of images carrying subjective scores of image quality based on the original infrared image dataset; A model training module for training a machine learning model by using each training image in the training set and the corresponding subjective image quality scores to obtain an initial quality evaluation model; a sub-dimension score calculation module, configured to determine the objective sub-dimension evaluation score of each test image by calculating the objective score of each objective quality index of each test image in the test set by using a no-reference evaluation algorithm; an objective evaluation module, which inputs the evaluation scores of each objective sub-dimension into the initial quality evaluation model, and obtains an objective image quality score of each test image; The model determination module is configured to determine the final image quality evaluation model according to the objective image quality scores and corresponding subjective image quality scores of each test image based on the initial quality evaluation model. 12. A device for evaluating infrared image quality, comprising: A model building module for obtaining an image quality evaluation model in advance using the infrared image quality evaluation method according to any one of claims 1 to 9; an image acquisition module for acquiring the infrared image to be evaluated; an objective scoring module, configured to determine the objective sub-dimension evaluation score of the infrared image to be evaluated by calculating the objective score of each objective quality index of the infrared image to be evaluated by using a reference-free evaluation algorithm; A quality evaluation module, configured to input the objective sub-dimension evaluation scores into the image quality evaluation model to obtain the image quality evaluation scores of the infrared images to be evaluated. 13. An electronic device, characterized in that it comprises a processor and a memory, and the processor is used to implement any one of claims 1 to 9 and/or as claimed in claim 10 when the processor is used to execute the computer program stored in the memory. Describe the steps of the infrared image quality evaluation method. 14. A readable storage medium, characterized in that, a computer program is stored on the readable storage medium, and when the computer program is executed by a processor, any one of claims 1 to 9 and/or as claimed in claims 10. Steps of the infrared image quality evaluation method.

Images