CN114881877A - Noise reduction method based on image airspace - Google Patents

Abstract

The invention discloses a noise reduction method based on an image airspace, which comprises the following steps: setting a filter window of image data to be processed, and respectively subtracting a central point value of the filter window and effective neighborhood point values around the central point value and taking an absolute value to correspondingly form a filter set; obtaining an HVS degree by using an HVS degree evaluation function; inputting the maximum point value of the filtering window into a local texture evaluation function for processing to obtain a local texture evaluation degree; carrying out adaptive threshold function processing on the maximum value and the minimum value in the set to obtain an adaptive threshold, and carrying out adaptive weight function processing on each element in the set according to the adaptive threshold to obtain the weight of a corresponding point of each element; and calculating to obtain the noise-reduced image estimation value by using a weighted summation mode. The noise reduction method based on the image airspace reduces the operation processing amount of image noise reduction, can rapidly remove noise, and reduces the requirements of related hardware equipment.

Description

A Noise Reduction Method Based on Image Spatial Domain

technical field

The invention relates to the technical field of image noise suppression, in particular to a noise reduction method based on image space.

Background technique

In the process of image acquisition, different levels of noise will be introduced with the sensor and its external environment, especially when the light source is insufficiently illuminated, which will cause the image signal-to-noise ratio to decrease and the detailed texture to be lost, thereby affecting the image quality and making the final display. Image effects affect the look and feel, and even submerge target signals in medical images, making it difficult to identify lesions. The relevant international standard organizations (ISO, CPIQ) also define the specific process of noise quantification. One of the comprehensive noise indicators that can represent the image observed by the human eye is visual noise vn (visual noise). The sensitivity difference between the grayscale signal and the color signal of the image. The general method of noise reduction algorithm design for vn is to filter the noise in the transform domain according to the CSF (Contrast Sensitivity Functions) characteristics of the human eye, and then restore the filtered signal through the inverse transform operation; but this The calculation process of the method is often complex, time-consuming, and difficult to implement in hardware. In the fields with strict bandwidth and delay requirements, more resources and costs need to be invested.

SUMMARY OF THE INVENTION

In order to solve the problems in the prior art, the present invention provides a noise reduction method based on image space, and the technical solution is as follows:

The present invention provides a noise reduction method based on image space, comprising the following steps:

S1. Perform filter window settings on the image data to be processed, denote the central point value of the filtering window as P _c , and denote the valid neighborhood point values around the central point value as P ₀ , P ₁ , and P ₂ , respectively. _{, . _ _ _} The maximum and minimum values are denoted as _Dmax and _Dmin , respectively;

S2, according to the central point value P _c of the filtering window, utilize the HVS degree evaluation function to obtain the HVS degree, denoted as d ₁ ;

S3. Design _a local texture evaluation function according to the HVS degree d1 and the preset noise reduction threshold parameter, and input the maximum point value of the filtering window into the local texture evaluation function for processing, so as to obtain the local texture evaluation degree , denoted as d ₂ ;

S4, according to the local texture evaluation degree d ₂ , perform adaptive threshold function processing on the maximum value and the minimum value in the set to obtain an adaptive threshold;

S5, according to the adaptive threshold, perform adaptive weight function processing on each element in the set to obtain the weight of the corresponding point of each element;

S6. Calculate the estimated value of the image after noise reduction by means of weighted summation.

Further, according to different observation conditions, adjust the noise reduction threshold parameter accordingly; the noise reduction threshold parameter includes the first noise reduction threshold parameter, the second noise reduction threshold parameter and the third noise reduction threshold parameter, which are respectively denoted as E ₁ , E ₂ , E ₃ , the first noise reduction threshold parameter is positively correlated with the maximum brightness of the environment where the display device is located, the second noise reduction threshold parameter is positively correlated with the maximum brightness output by the display device, and the third noise reduction threshold parameter is positively correlated with the maximum brightness output by the display device. The threshold parameter is positively related to the viewing angle of the image viewer per pixel of the display device.

Further, in step S2, the formula of the HVS degree evaluation function is as follows:

When L≤2 ^Depth-1 , d ₁ =2 ^1-Depth ×L

When L>2 ^Depth-1 , d ₁ =2-2 ^1-Depth ×L

Among them, Depth is the bit depth of the image data, and L is the brightness value of the center point.

Further, in step S3, the formula of the local texture evaluation function is as follows:

When D _max ≤ E ₃ , d ₂ =1;

When E ₃ <D _max ≤Th _end ,

When D _max >Th _end , d ₂ =0;

Wherein, Th _end =E ₁ -E ₂ ×d ₁ +m ₁ , where m ₁ is an image noise level parameter.

Further, in step S4, the formula of the adaptive threshold function is as follows:

When d ₂ =0,

When 0<d ₂ <1,

When d ₂ =1, Th _low =Th _high =D _max .

Further, in step S5, the formula of the adaptive weight function is as follows:

When 0≤D _i <Th _low , w _i =1

When Th _low ≤ D _i <Th _high ,

When D _i >Th _high , w _i =0

Wherein, D _i is the element in the set D, _wi is the weight of the corresponding point of P _i , i=0, 1, 2...n.

Further, in step S6, the calculation formula of the image estimated value after the noise reduction of the center point value of the filtering window is as follows:

Among them, _{Pi is the effective neighborhood point value around the center point P c ,} and Output is the image estimated value.

Further, a filter window corresponds to a filter cycle, and according to the local texture evaluation degree in the current filter cycle and the local texture evaluation degree in the previous filter cycle, a local noise evaluation function is designed, and the maximum point value of the filter window is input to The local noise evaluation function is processed to obtain the local noise evaluation degree, denoted as d ₃ ;

The local noise evaluation degree in the previous filtering cycle is taken as the image noise level parameter in the current filtering cycle to calculate the local texture evaluation degree in the current filtering cycle.

Further, the formula of the local noise evaluation function is as follows:

d ₃ =d ₂ ×D _max +(1−d ₂ )×m ₂

where m ₂ is the local texture evaluation degree calculated in the previous filtering cycle.

Further, when the central point value P _c is 0 or the maximum value, the HVS degree d ₁ is 0, and the function shape of the HVS degree evaluation function is a Gaussian curve; the function shape of the local texture evaluation function is a trapezoidal curve. .

The beneficial effects brought by the technical scheme provided by the invention are as follows:

a. It reduces the computational processing amount of image noise reduction, can quickly denoise, and reduces the requirements of related hardware equipment;

b. It can adapt to the original image data of various sensors, and has better compatibility and flexibility.

Description of drawings

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

FIG. 1 is a schematic flowchart of a noise reduction method based on image space provided by an embodiment of the present invention;

2 is a schematic diagram of an HVS degree evaluation function in a noise reduction method based on image space provided by an embodiment of the present invention;

3 is a schematic diagram of a local texture evaluation function in an image space-based noise reduction method provided by an embodiment of the present invention;

FIG. 4 is a schematic diagram of a filtering window of a noise reduction method based on an image space domain provided by an embodiment of the present invention.

Detailed ways

In order to make those skilled in the art better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only Embodiments are part of the present invention, but not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

It should be noted that the terms "first", "second" and the like in the description and claims of the present invention and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific sequence or sequence. It is to be understood that the data so used may be interchanged under appropriate circumstances such that the embodiments of the invention described herein can be practiced in sequences other than those illustrated or described herein. Furthermore, the terms "comprising" and "having", and any variations thereof, are intended to cover non-exclusive inclusion, for example, a process, method, apparatus, product or device comprising a series of steps or units is not necessarily limited to those expressly listed Rather, those steps or units may include other steps or units not expressly listed or inherent to these processes, methods, products or devices.

In an embodiment of the present invention, a noise reduction method based on image space is provided, referring to FIG. 1 , including the following steps:

S1, filter window settings are performed on the image data to be processed, one filter window corresponds to one filter cycle, the center point value of the filter window is denoted as P _c , and the effective neighborhood point values around the center point value are respectively denoted as _{P 0} , _{P 1} , _{P 2} , _. and the minimum value are denoted as _Dmax and _Dmin respectively;

S2, according to the central point value P _c of the filtering window, utilize the HVS degree evaluation function to obtain the HVS degree, denoted as d ₁ ;

Among them, the HVS degree evaluation function calculates the HVS degree d ₁ through P _c , and the value of d ₁ ranges from 0 to 1. Since the human eye perceives the texture and noise of the middle gray level more obviously than the too dark and too bright areas, Therefore, when the value of P _c is close to the middle, the HVS degree evaluation function achieves the maximum value. When the value of P _c is 0 or the maximum value, the HVS degree is 0. A typical function form of the HVS degree evaluation function is as shown in Figure 2. The Gaussian curve shown.

S3. Design _a local texture evaluation function according to the HVS degree d1 and the preset noise reduction threshold parameter, and input the maximum point value of the filtering window into the local texture evaluation function for processing, so as to obtain the local texture evaluation degree , denoted as d ₂ ;

Wherein, according to different observation conditions, for example, the external observation environment such as the distance between the observer and the display, the image display size, the observation angle, etc., the noise reduction threshold parameter is adjusted accordingly; the noise reduction threshold parameter includes the first noise reduction threshold parameter, the second noise reduction threshold parameter The noise reduction threshold parameter and the third noise reduction threshold parameter are respectively denoted as E ₁ , E ₂ , and E ₃ , the first noise reduction threshold parameter is positively correlated with the maximum brightness of the environment where the display device is located, and the second noise reduction threshold parameter is positively correlated. The threshold parameter is positively correlated with the maximum brightness output by the display device. E ₁ and E ₂ are the degree of participation of the HVS degree in the algorithm. The two jointly restrict the participation of the image highlights in the calculation of d _2. When the ambient light brightness obviously exceeds the maximum brightness of the display, More bright pixels in the image participate in the calculation of _d2 . The third noise reduction threshold parameter is positively correlated with the viewing angle of the image observer to the unit pixel of the display device, which corresponds to the opening angle of the unit pixel of the display to the human eye, which can be calculated by dividing the pixel size of the display by the viewing distance. The noise reduction threshold parameter can be preset manually, or the AI camera can be used to identify and detect the observer and the display device in real time, and the current observation condition can be obtained through an algorithm to adjust the noise threshold parameter accordingly, so as to realize the automatic adjustment of the parameters.

The value range of the local texture evaluation degree d ₂ is also 0 to 1. The smaller D _max is, the larger the value of d ₂ is. A typical function shape of the local texture evaluation function is a trapezoidal function as shown in Figure 3. Specifically, The formula of the local texture evaluation function is as follows:

When D _max ≤ E ₃ , d ₂ =1;

When E ₃ <D _max ≤Th _end ,

When D _max >Th _end , d ₂ =0;

In the formula, Th _end =E ₁ -E ₂ ×d ₁ +m ₁ , and m ₁ is an image noise level parameter.

It should be noted that there are two modes for the value of m ₁ , which correspond to the two modes of the noise reduction method, namely, the FIR (Finite Impulse Response) mode and the IIR (Infinite Impulse Response) mode. In FIR mode, the filtering of the previous window does not affect the filtering of the latter window, so m ₁ is input as a fixed parameter to further reduce the amount of computation.

In the IIR mode, the filtering of the previous window will affect the filtering of the latter window, so the front and back feedback mechanism is established through m _1. In this mode, the noise and local texture evaluation of the algorithm is more accurate, and the noise reduction effect is better; Specifically, According to the local texture evaluation degree in the current filtering cycle and the local texture evaluation degree in the previous filtering cycle, a local noise evaluation function is designed, and the maximum point value of the filtering window is input into the local noise evaluation function for processing to obtain The local noise evaluation degree, denoted as d ₃ , the formula of the local noise evaluation function is as follows:

d ₃ =d ₂ ×D _max +(1−d ₂ )×m ₂

where m ₂ is the local texture evaluation degree calculated in the previous filtering cycle.

The local noise evaluation degree in the previous filtering cycle is taken as the image noise level parameter m ₁ in the current filtering cycle to calculate the local texture evaluation degree in the current filtering cycle. In the first filtering cycle, the image noise level parameter m ₁ uses a preset initial value.

S4, according to the local texture evaluation degree d ₂ , perform adaptive threshold function processing on the maximum value and the minimum value in the set to obtain an adaptive threshold;

Wherein, a dual adaptive threshold setting method is adopted to obtain the adaptive thresholds Th _{l ow} and Th _high , and the formula of the adaptive threshold function is as follows:

When d ₂ =0,

When 0<d ₂ <1,

When d ₂ =1, Th _low =Th _high =D _max .

S5, according to the adaptive threshold, perform adaptive weight function processing on each element in the set to obtain the weight of the corresponding point of each element;

Wherein, the formula of the adaptive weight function is as follows:

When 0≤D _i <Th _low , w _i =1

When Th _low ≤ D _i <Th _high ,

When D _i >Th _high , w _i =0

Wherein, D _i is the element in the set D, _wi is the weight of the corresponding point of P _i , i=0, 1, 2...n.

S6. Calculate the estimated value of the image after denoising by means of weighted summation.

Wherein, the calculation formula of the image estimated value after the noise reduction of the center point value of the filtering window is as follows:

Among them, _{Pi is the effective neighborhood point value around the center point P c ,} and Output is the image estimated value.

After step S6, it also includes data post-processing, including demosaic, gamma, sharpening and other processes.

Referring to FIG. 1 , this embodiment utilizes an HVS level estimator, a local texture level estimator, a local noise level estimator, and a dual adaptive threshold weight calculator to process corresponding function algorithms. Among them, the local noise level estimator can only be activated and enabled in this algorithm module in IIR mode. In each filter window, using the above evaluator, calculate d ₁ through the HVS degree evaluation function, and calculate d ₂ through the local texture evaluation function. Adjust the appropriate noise reduction threshold parameters according to the eye observation angle. In the spatial filtering, the current filtering noise coefficient is calculated by d ₁ , d ₂ , E ₁ , E ₂ , and E ₃ to obtain the local noise evaluation degree d ₃ . Since the image has continuity within a certain range, d ₂ , d ₃ is used as feedback data to the local texture level estimator and the local noise level estimator in the next filtering process. In this case, the filtering algorithm can be changed from FIR filter to IIR filter, and finally the double adaptive threshold is calculated to obtain the weight calculation function, the weight of each element in the filter window is calculated by the difference between each element in the filter window and the center, and the denoising estimate value of the target position is obtained by weighted summation.

Compared with other noise reduction methods, such as cosine transform, wavelet transform, bilateral filtering, nlm, etc., this embodiment comprehensively considers the characteristics of HVS, local texture characteristics, and local noise levels, and can be adjusted according to actual application conditions. Switching the algorithm to IIR or FIR is more flexible; compared with methods such as transform domain noise reduction, the noise reduction method of this embodiment operates directly in the spatial domain, which reduces the computational complexity and is more conducive to hardware implementation of the algorithm. For specific industries, such as surgical imaging display systems in the medical industry, the system can also meet strict requirements on delay and bandwidth; for HVS noise perception conditions are known, the distance between the display device and the image observer, observation Compared with other noise reduction methods, inputting these necessary parameters into the noise reduction unit can obtain better image quality for visual noise perception; in addition, the dual adaptive threshold weight calculator can also make the image It retains better edge and texture details, and the principle is similar to the edge-preserving feature of bilateral filtering.

The following is an example of a 5×5 filter window of size:

(1) The input image data is the original raw data, the filter window size is 5×5, the center point value is P _c , see Figure 4, the set D is {|P _c -P _i |} _{i=0,1,2. ..n} , see Figure 4, when the center pixel is R, B channel, n=7, when the center pixel is G channel, n=11; D _max is the maximum value of the set D, and the selected working mode is IIR mode .

(2) Calculate the degree of HVS d ₁

The formula of the HVS degree evaluation function is as follows:

When L≤2 ^Depth-1 , d ₁ =2 ^1-Depth ×L

When L>2 ^Depth-1 , d ₁ =2-2 ^1-Depth ×L

Among them, Depth is the bit depth of the image data, and L is the brightness value of the center point. Since the human eye is relatively sensitive to the G channel, when P _c is in the G channel, L=P _c , otherwise, L is obtained according to the demosaic algorithm, for example, L is obtained by interpolation according to the Gr and Gb channels around the center point. .

(3) Calculate the local texture evaluation degree d ₂

When D _max ≤ E ₃ , d ₂ =1;

When E ₃ <D _max ≤Th _end ,

When D _max >Th _end , d ₂ =0;

In the formula, Th _end =E ₁ -E ₂ ×d ₁ +m ₁ , m ₁ is the image noise level parameter, in the IIR mode, m ₁ is valued by the parameter passed by the local noise level estimator in the previous filtering cycle. E ₁ and E ₂ are the degree of participation of HVS degree d ₁ in the algorithm, and E ₃ is positively related to the visual angle of the image observer to the unit pixel.

(4) Calculate and update the local noise evaluation degree d ₃ :

The formula of the local noise evaluation function is as follows:

d ₃ =d ₂ ×D _max +(1−d ₂ )×m ₂

where m ₂ is the local texture evaluation degree calculated in the previous filtering cycle.

(5) Calculate the adaptive threshold by _d2

(6) The weight wi corresponding to each pixel P _i in the filtering window is obtained through the double adaptive threshold weight calculator. The estimated value of the denoised image is obtained according to the following formula:

Among them, Output is the image estimation value.

The noise reduction method based on the image space domain provided by the present invention performs spatial filtering and noise reduction based on the analysis of hvs and texture features. Due to the processing in the spatial domain, the amount of calculation is reduced, and the method can also be applied to the original raw data of the sensor and grayscale. There are two cases of degree data, which has better compatibility and flexibility. The noise reduction method based on image space provided by the present invention can be applied to imaging devices, imaging modules, ISP (Image Signal Processor), DSP (Digital Signal Processor) and display devices.

The above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection of the present invention. within the range.

Claims (10)

1. A noise reduction method based on an image space domain is characterized by comprising the following steps:

s1, setting a filter window of the image data to be processed, and recording the central point value of the filter window as P _c Respectively recording effective neighborhood point values around the central point value as P ₀ ，P ₁ ，P ₂ ，…，P _n And correspondingly form a set { | P _c -P _i |} _{i＝0,1,2...n} Denoted D, D _i For each element in the set, the maximum and minimum values in the set are respectively denoted as D _max And D _min ；

S2, according to the central point value P of the filtering window _c Using the HVS degree evaluation function to obtain the HVS degree, which is recorded as d ₁ ；

S3, according to the HVS degree d ₁ And a predetermined noise reduction threshold parameter, design officePartial texture evaluation function, inputting the maximum point value of the filtering window into the partial texture evaluation function for processing to obtain partial texture evaluation degree, which is recorded as d ₂ ；

S4, evaluating degree d according to the local texture ₂ Carrying out adaptive threshold function processing on the maximum value and the minimum value in the set to obtain an adaptive threshold;

s5, according to the adaptive threshold, performing adaptive weight function processing on each element in the set to obtain the weight of the corresponding point of each element;

and S6, calculating to obtain the noise-reduced image estimation value by using a weighted summation mode.

2. The image-space-domain-based noise reduction method according to claim 1, wherein the noise reduction threshold parameter is adjusted according to different observation conditions; the noise reduction threshold parameter comprises a first noise reduction threshold parameter, a second noise reduction threshold parameter and a third noise reduction threshold parameter, which are respectively marked as E ₁ 、E ₂ 、E ₃ The first noise reduction threshold parameter is positively correlated with the maximum brightness of the environment where the display device is located, the second noise reduction threshold parameter is positively correlated with the maximum brightness output by the display device, and the third noise reduction threshold parameter is positively correlated with the visual field angle of an image observer to the unit pixel of the display device.

3. The image-spatial-domain-based noise reduction method according to claim 1, wherein in step S2, the HVS degree evaluation function is formulated as follows:

when L is less than or equal to 2 ^Depth-1 When d is greater than ₁ ＝2 ^1-Depth ×L

When L > 2 ^Depth-1 When d is greater than ₁ ＝2-2 ^1-Depth ×L

Where Depth is the bit Depth of the image data, and L is the central point brightness value.

4. The image-spatial-domain-based noise reduction method according to claim 2, wherein in step S3, the formula of the local texture evaluation function is as follows:

when D is present _max ≤E ₃ When d is greater than ₂ ＝1；

When E is ₃ <D _max ≤Th _end When the temperature of the water is higher than the set temperature,

when D is present _max >Th _end When d is greater than ₂ ＝0；

Wherein Th _end ＝E ₁ -E ₂ ×d ₁ +m ₁ Wherein m is ₁ Is an image noise level parameter.

5. The method for noise reduction based on image spatial domain according to claim 1, wherein in step S4, the adaptive threshold function is formulated as follows:

when d is ₂ When the content is equal to 0, the content,

when 0 is present<d ₂ <When the pressure of the mixture is 1, the pressure is lower,

when d is ₂ When 1, Th _low ＝Th _high ＝D _max 。

6. The method for noise reduction based on image spatial domain according to claim 5, wherein in step S5, the formula of the adaptive weight function is as follows:

when D is more than or equal to 0 _i <Th _low When w _i ＝1

When Th is _low ≤D _i <Th _high When the temperature of the water is higher than the set temperature,

when D is present _i >Th _high When w _i ＝0

Wherein D is _i As elements in the set D, w _i Is P _i The weight of the corresponding point, i ═ 0,1,2.. n.

7. The method for noise reduction based on image spatial domain according to claim 6, wherein in step S6, the formula for calculating the noise-reduced image estimation values of the center point values of the filtering window is as follows:

wherein Output is an image estimation value.

8. The image-space-domain-based noise reduction method according to claim 1, wherein one filtering window corresponds to one filtering cycle, a local noise evaluation function is designed according to the local texture evaluation degree in the current filtering cycle and the local texture evaluation degree in the previous filtering cycle, the maximum point value of the filtering window is input to the local noise evaluation function for processing to obtain a local noise evaluation degree, which is denoted as d ₃ ；

And taking the local noise evaluation degree in the previous filtering period as an image noise level parameter in the current filtering period to calculate the local texture evaluation degree in the current filtering period.

9. The method of claim 8, wherein the local noise evaluation function is formulated as follows:

d ₃ ＝d ₂ ×D _max +(1-d ₂ )×m ₂

wherein m is ₂ The degree is evaluated for the local texture calculated in the previous filtering cycle.

10. Image spatial domain based noise reduction according to claim 1Method, characterized in that, when the central point value P is _c At 0 or maximum, the HVS degree d ₁ The function form of the HVS degree evaluation function is a Gaussian curve; the function form of the local texture evaluation function is a trapezoidal curve.

Images