CN108460751A - The 4K ultra high-definition image quality evaluating methods of view-based access control model nerve - Google Patents
The 4K ultra high-definition image quality evaluating methods of view-based access control model nerve Download PDFInfo
- Publication number
- CN108460751A CN108460751A CN201710098240.3A CN201710098240A CN108460751A CN 108460751 A CN108460751 A CN 108460751A CN 201710098240 A CN201710098240 A CN 201710098240A CN 108460751 A CN108460751 A CN 108460751A
- Authority
- CN
- China
- Prior art keywords
- video
- frame
- access control
- ultra high
- control model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 21
- 210000005036 nerve Anatomy 0.000 title claims abstract description 13
- 238000010586 diagram Methods 0.000 claims abstract description 9
- 238000009499 grossing Methods 0.000 claims description 3
- 238000011156 evaluation Methods 0.000 abstract description 5
- 230000000007 visual effect Effects 0.000 description 7
- 230000008859 change Effects 0.000 description 3
- 230000010365 information processing Effects 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 230000008844 regulatory mechanism Effects 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 210000004556 brain Anatomy 0.000 description 1
- 239000012141 concentrate Substances 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000005284 excitation Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000008450 motivation Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000013441 quality evaluation Methods 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 238000010561 standard procedure Methods 0.000 description 1
- 230000026676 system process Effects 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 230000002747 voluntary effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/46—Extracting features or characteristics from the video content, e.g. video fingerprints, representative shots or key frames
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30168—Image quality inspection
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of 4K ultra high-definition image quality evaluating methods of view-based access control model nerve, include the following steps：Step 1, original video and video to be measured are respectively divided into several frame images；Step 2, the notable figure in each frame image of video to be measured is calculated；Step 3, the structure similar diagram of each frame in original video is calculated；Step 4, the notable figure of corresponding frame and structure similar diagram are done into weighted comprehensive, calculates the picture quality of video to be measured.The method of the present invention has the characteristics that speed is fast, expense is low and can be embedded into video system itself, and vision significance algorithm is applied in video evaluation.
Description
Technical field
The present invention relates to method for evaluating video quality, more specifically to a kind of 4K ultra high-definitions of view-based access control model nerve
Image quality evaluating method.
Background technology
Vision attention is a kind of significant psychological regulation mechanism in human visual system's process of information processing, and people is allowed to select
The notable feature for obtaining to selecting property institute's observation object, to substantially reduce information processing capacity.Vision attention is in computer vision
Play important role.Many tasks of computer vision, such as scene analysis, object detection, video tracking, object are known
Not, retrieval, estimation and image recovery etc., all studied and improve performance using vision attention.
What you were seen is that you want to see, this is a main Research foundation of vision significance.At any time, ring
The visual information that border is showed can be handled considerably beyond human eye.Vision significantly allow people choose with instantly oneself into
The relevant information of capable behavior.In order to adapt to this potential burden, human brain also has a series of corresponding conspicuousness mechanism.
Main includes two aspects：It is conspicuousness first to be used for choosing relevant information and ignoring incoherent information and can use
To speculate information.In addition conspicuousness can adjust according to the target of behavior and enhance selected information.The research of conspicuousness
It can be divided into many types according to different situations.And time and space significance is wherein than major one kind.
When we observe ambient enviroment, often due to attention is had selection by behavior purpose demand or local scene clue
Ground concentrates on some or certain scenery, to select the representative of certain point or region as scenery.The sheet of vision attention
Matter is a psychological regulation mechanism of human vision, and the mankind choose interested area from the bulk information of external world's input whereby
Domain, and the notable feature of target of interest is selectively obtained to a certain extent, to reduce information processing capacity.
Conspicuousness is introduced into image quality evaluation and has been obtained for studying.Due to the introducing of vision significance, obtain more
Meet the evaluating objective quality result of human eye.But vision significance algorithm is primarily directed to image at present, and video is special
It is that ultra high-definition video wants the more of complexity compared to image, cannot directly applies mechanically the conspicuousness algorithm for image, because the third dimension
Temporal signatures and movable information need to be considered.
Invention content
The object of the present invention is to provide a kind of 4K ultra high-definition image quality evaluating methods of view-based access control model nerve, solve existing
Vision significance algorithm is used only for picture appraisal in technology, and the problem of cannot be used for video evaluation.
To achieve the above object, the present invention adopts the following technical scheme that：
A kind of 4K ultra high-definition image quality evaluating methods of view-based access control model nerve, include the following steps：Step 1, original to regard
Frequency and video to be measured are respectively divided into several frame images；Step 2, the notable figure in each frame image of video to be measured is calculated；Step 3,
Calculate the structure similar diagram of each frame in original video；Step 4, that the notable figure of corresponding frame and structure similar diagram are done weighting is comprehensive
It closes, calculates the picture quality of video to be measured.
Preferably, step 2 further comprises：Step 2.1, video is formed by continuous frame by one and is decomposed into 3 significantly
Property body Ci(i=1,2,3) corresponds to 3 kinds of different features, brightness, color and movement respectively；Step 2.2, each conspicuousness body
Different scale j is resolved into progress, establishes a series of Gauss tower C={ Cij, wherein i=1,2,3, j=1 ..., L；Step
2.3, energy function E is minimized, it includes a data item EdWith smoothing factor Es, wherein：E (C)=λd·Ed(C)+λs·Es
(C), wherein λdAnd λsFor corrected parameter；Step 2.4, last conspicuousness is that each scale and feature are average：Wherein, j=1 ..., L.
Preferably, step 3 further comprises：Enable VoAnd VdOriginal and distortion video is indicated respectively, and dimension is M*N*F；G and f
V is indicated respectivelydA frame image.For the structure chart SS of calculated distortion videof, g and f are divided into many fritter x and y, x and y
The calculating of SSIM values is as follows：Wherein μx, μyAnd σx, σyIt is the equal of x and y
Value and variance, σxyIt is the covariance of x and y, C1、C2It is constant.The SSIM values of all small image blocks in one frame image constitute
Structure chart SSf(f=1 ..., F).
Preferably, step 4 further comprises：Quality per frame
Preferably, final video quality value VQ is：
In the above-mentioned technical solutions, the 4K ultra high-definition image quality evaluating methods of view-based access control model nerve of the invention have speed
The characteristics of degree is fast, expense is low and can be embedded into video system itself, and vision significance algorithm is applied to video evaluation
In.
Description of the drawings
Fig. 1 is the flow chart of the method for the present invention.
Specific implementation mode
The technical solution further illustrated the present invention with reference to the accompanying drawings and examples.
The method of the present invention is based on 2 important vision significance principles, i.e. spatial domain conspicuousness and time domain conspicuousness.
Spatial domain conspicuousness：In the task of any visual search, the mode that human eye is watched attentively can be divided into inherence and external two
Kind.Inherence pays attention to assuming to think that object controls vision attention.This is also referred to as from top to bottom, the vision attention of target excitation.
It is a kind of voluntary, needs the process made great efforts.In addition vision attention can be automatically driven to by external driving by extrinsic motivation
Some position.This is referred to as the vision attention for encouraging driving from down to up.Flashing lamp on highway is regarded from external driving
Feel attention.External vision attention automatically drives note that response process is rapid.Vision attention external from down to up includes very
The driving attribute of multiple types.For example, in visual search, spatial cues and the setting of unexpected vision can cause vision attention.
Other visual signatures such as single feature (the vertical target in red objects or horizontal in green color background), can
Effectively arouse attention.
Time domain conspicuousness：It is the feature of vision selection in time.The input of visual signal is constantly to change at any time
, observer needs extraction and the relevant information of behavior from vision inlet flow.A kind of standard technique of research time domain conspicuousness
It is that object sequence is played with the speed of 20 objects per second.This processing method and selection mechanism can allow researcher to find to regard
Feel the video playout speed that information can be extracted.
In view of this, the present invention has used for reference the dense time and space significance of Konstantinos Rapantzikos et al. propositions
Model (Dense Spatiotemporal Salient Model), abbreviation KR models are the conspicuousness moulds suitable for video
One of type.The model characterizes video using multiple dimensioned body, and space-time operation is carried out on three dimensions.The calculating of conspicuousness is logical
The process of a global minimization is crossed to realize.The constraints of the minimum is related to a series of visual signature information
, including spatial domain proximity, scale and characteristic similarity (brightness, color, movement).The extreme value of above-mentioned conspicuousness response is just chosen
For distinguishing feature, it has therefore proved that this method has reached good balance between brightness and information.
Based on above-mentioned principle, commented as shown in Figure 1, the present invention discloses a kind of 4K ultra high-definition picture qualities of view-based access control model nerve
Valence method comprising key step below：
S1：Original video and video to be measured are respectively divided into several frame images.
S2：Calculate the notable figure in each frame image of video to be measured.
S3：Calculate the structure similar diagram of each frame in original video.
S4：The notable figure of corresponding frame and structure similar diagram are done into weighted comprehensive, calculate the picture quality of video to be measured.
Assuming that V represents a video being made of continuous frame, q=(x, y, t) represents an individual moment point.So
Q means that the element of volume in an individual.V (q) is enabled to indicate pixel values of the V at q points.At this point, the main calculating of the notable figure of S2
Step includes：
S2.1：Video is formed by continuous frame by one and is decomposed into 3 conspicuousness body Ci(i=1,2,3) corresponds to 3 respectively
Kind different feature, brightness, color and movement；
S2.2：Different scale j is resolved into each conspicuousness body progress, establishes a series of Gauss tower C={ Cij,
Middle i=1,2,3, j=1 ..., L；
S2.3：Energy function E is minimized, it includes a data item EdWith smoothing factor Es, wherein：E (C)=λd·Ed
(C)+λs·Es(C), wherein λdAnd λsFor corrected parameter；
S2.4：Last conspicuousness is that each scale and feature are average：Wherein, j=1 ..., L.
In above-mentioned steps, it is assumed that the result of the first scale is chosen as final Saliency maps, i.e., such as enables j=1, then：
Obviously, if certain part of video image is relatively more prominent, human eye is also relatively more to their concern, then managing institute
This certain partial content weight shared in the evaluation result of video also should be big.Therefore, salient region is determining entirety
Video evaluation value in should occupy an important position.Based on such idea, the present invention is parallel to execute while executing S2
S3。
Enable VoAnd VdOriginal and distortion video is indicated respectively, and dimension is M*N*F.G and f indicate V respectivelydA frame image.For
The structure chart SS of calculated distortion videof, the calculating for the SSIM values that g and f are divided into many fritter x and y, x and y is as follows：
Wherein, μx, μyAnd σx, σyIt is the mean value and variance of x and y, σxyIt is the covariance of x and y, C1、C2It is constant.One frame figure
The SSIM values of all small image blocks as in constitute structure chart SSf(f=1 ..., F).
Finally, use notable figure as the weighted factor of structure similar diagram, and the final video mass value of entire video is each
Frame weighted graph is averaged, specifically：
Quality per frame
According to above-mentioned it is assumed that assuming that the result of the first scale is chosen as final Saliency maps：
At this timeThen：
Quality per frame
Final video quality value VQ is：
Those of ordinary skill in the art it should be appreciated that more than embodiment be intended merely to illustrate the present invention,
And be not used as limitation of the invention, as long as in the spirit of the present invention, the change to embodiment described above
Change, modification will all be fallen within the scope of claims of the present invention.
Claims (5)
1. a kind of 4K ultra high-definition image quality evaluating methods of view-based access control model nerve, which is characterized in that include the following steps：
Step 1, original video and video to be measured are respectively divided into several frame images；
Step 2, the notable figure in each frame image of video to be measured is calculated；
Step 3, the structure similar diagram of each frame in original video is calculated；
Step 4, the notable figure of corresponding frame and structure similar diagram are done into weighted comprehensive, calculates the picture quality of video to be measured.
2. the 4K ultra high-definition image quality evaluating methods of view-based access control model nerve as described in claim 1, which is characterized in that step
2 further comprise：
Step 2.1, video is formed by continuous frame by one and is decomposed into 3 conspicuousness body Ci(i=1,2,3) corresponds to 3 kinds respectively
Different feature, brightness, color and movements；
Step 2.2, different scale j is resolved into each conspicuousness body progress, establishes a series of Gauss tower C={ Cij,
Middle i=1,2,3, j=1 ..., L；
Step 2.3, energy function E is minimized, it includes a data item EdWith smoothing factor Es, wherein：
E (C)=λd·Ed(C)+λs·Es(C), wherein λdAnd λsFor corrected parameter；
Step 2.4, last conspicuousness is that each scale and feature are average：
Wherein, j=1 ..., L.
3. the 4K ultra high-definition image quality evaluating methods of view-based access control model nerve as claimed in claim 2, which is characterized in that step
3 further comprise：
Enable VoAnd VdOriginal and distortion video is indicated respectively, and dimension is M*N*F；G and f indicate V respectivelydA frame image.In order to count
Calculate the structure chart SS of distortion videof, the calculating for the SSIM values that g and f are divided into many fritter x and y, x and y is as follows：
Wherein μx, μyAnd σx, σyIt is the mean value and variance of x and y, σxyIt is the covariance of x and y, C1、C2It is constant.In one frame image
The SSIM values of all small image blocks constitute structure chart SSf(f=1 ..., F).
4. the 4K ultra high-definition image quality evaluating methods of view-based access control model nerve as claimed in claim 3, which is characterized in that step
4 further comprise：
Quality per frame
5. the 4K ultra high-definition image quality evaluating methods of view-based access control model nerve as claimed in claim 4, which is characterized in that final
Video quality value VQ be：
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710098240.3A CN108460751A (en) | 2017-02-22 | 2017-02-22 | The 4K ultra high-definition image quality evaluating methods of view-based access control model nerve |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710098240.3A CN108460751A (en) | 2017-02-22 | 2017-02-22 | The 4K ultra high-definition image quality evaluating methods of view-based access control model nerve |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108460751A true CN108460751A (en) | 2018-08-28 |
Family
ID=63220850
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710098240.3A Pending CN108460751A (en) | 2017-02-22 | 2017-02-22 | The 4K ultra high-definition image quality evaluating methods of view-based access control model nerve |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108460751A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110312124A (en) * | 2019-07-31 | 2019-10-08 | 中国矿业大学 | A quality correction method for mobile inspection video based on salient multi-feature fusion |
| CN111385567A (en) * | 2020-03-12 | 2020-07-07 | 上海交通大学 | Ultra-high-definition video quality evaluation method and device |
| CN114373085A (en) * | 2021-12-31 | 2022-04-19 | 江苏任务网络科技有限公司 | Calculation Method of Image Similarity Based on Neighborhood Similarity |
| CN115239647A (en) * | 2022-07-06 | 2022-10-25 | 杭州电子科技大学 | Full-reference video quality evaluation method based on two stages of self-adaptive sampling and multi-scale time sequence |
| CN117934354A (en) * | 2024-03-21 | 2024-04-26 | 共幸科技(深圳)有限公司 | Image processing method based on AI algorithm |
-
2017
- 2017-02-22 CN CN201710098240.3A patent/CN108460751A/en active Pending
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110312124A (en) * | 2019-07-31 | 2019-10-08 | 中国矿业大学 | A quality correction method for mobile inspection video based on salient multi-feature fusion |
| CN110312124B (en) * | 2019-07-31 | 2020-09-08 | 中国矿业大学 | A mobile inspection video quality correction method based on saliency multi-feature fusion |
| CN111385567A (en) * | 2020-03-12 | 2020-07-07 | 上海交通大学 | Ultra-high-definition video quality evaluation method and device |
| CN114373085A (en) * | 2021-12-31 | 2022-04-19 | 江苏任务网络科技有限公司 | Calculation Method of Image Similarity Based on Neighborhood Similarity |
| CN115239647A (en) * | 2022-07-06 | 2022-10-25 | 杭州电子科技大学 | Full-reference video quality evaluation method based on two stages of self-adaptive sampling and multi-scale time sequence |
| CN117934354A (en) * | 2024-03-21 | 2024-04-26 | 共幸科技(深圳)有限公司 | Image processing method based on AI algorithm |
| CN117934354B (en) * | 2024-03-21 | 2024-06-11 | 共幸科技(深圳)有限公司 | Image processing method based on AI algorithm |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Jin et al. | Pedestrian detection with super-resolution reconstruction for low-quality image | |
| Wang et al. | Multi-scale dilated convolution of convolutional neural network for crowd counting | |
| Song et al. | Deep sliding shapes for amodal 3d object detection in rgb-d images | |
| CN108460751A (en) | The 4K ultra high-definition image quality evaluating methods of view-based access control model nerve | |
| US8503770B2 (en) | Information processing apparatus and method, and program | |
| KR101333347B1 (en) | Image processing method and image processing device | |
| CN103065326B (en) | Target detection method based on time-space multiscale motion attention analysis | |
| CN111491187B (en) | Video recommendation method, device, equipment and storage medium | |
| CN112580576A (en) | Face spoofing detection method and system based on multiscale illumination invariance texture features | |
| EP2246807A1 (en) | Information processing apparatus and method, and program | |
| WO2011148562A1 (en) | Image information processing apparatus | |
| Ip et al. | Saliency-assisted navigation of very large landscape images | |
| CN111260738A (en) | Multi-scale target tracking method based on relevant filtering and self-adaptive feature fusion | |
| CN110827193A (en) | Panoramic video saliency detection method based on multi-channel features | |
| CN105701467A (en) | Many-people abnormal behavior identification method based on human body shape characteristic | |
| CN103247038B (en) | A kind of global image information synthesis method of visual cognition model-driven | |
| CN105513080B (en) | An Infrared Image Target Saliency Evaluation Method | |
| CN110781962A (en) | Target detection method based on lightweight convolutional neural network | |
| CN105303571A (en) | Time-space saliency detection method for video processing | |
| CN114067428A (en) | Multi-view multi-target tracking method and device, computer equipment and storage medium | |
| Chaabouni et al. | ChaboNet: Design of a deep CNN for prediction of visual saliency in natural video | |
| Singh et al. | Learning to Predict Video Saliency using Temporal Superpixels. | |
| Chen et al. | SRCBTFusion-Net: An efficient fusion architecture via stacked residual convolution blocks and transformer for remote sensing image semantic segmentation | |
| CN108460794A (en) | A kind of infrared well-marked target detection method of binocular solid and system | |
| CN110245660B (en) | Webpage glance path prediction method based on saliency feature fusion |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20180828 |
|
| WD01 | Invention patent application deemed withdrawn after publication |