A video quality assessment method based on just noticeable difference and blocking effect

Abstract

The present invention proposes a video quality evaluation method based on minimum perceptible difference and block effect, comprising the following steps: step S1: according to the change of pixel values ​​between adjacent blocks of a video frame and the visibility threshold of the human visual system to different background brightness values, the minimum perceptible difference of a pixel point is calculated; the perceptible difference is determined by whether the change of the video pixel point can be perceived by the human eye; step S2: judging the perceptible coding effect of each pixel point by the calculated minimum perceptible difference, and applying the judgment result to the detection of the block effect; step S3: detecting and extracting the significant area of ​​the video by using the frame difference method, and taking the significant area as a reference sample for evaluating the video quality; the present invention can realize real-time and accurate objective quality evaluation of the video.

Description

Technical Field

[0001] The invention relates to the technical field of video image quality evaluation, in particular to a video quality evaluation method based on a minimum perceptible difference and a block effect. Background Art

[0002] With the rapid development of monitoring system technology and the reduction of application costs, more and more monitoring systems are being used in the construction field to meet more security needs. Monitoring video will inevitably produce various types of distortion during the acquisition and transmission process, which will not only reduce people's subjective feelings when observing the video, but also affect the post-processing and application of the video, and even cause safety protection problems. For example, casualties and engineering quality accidents in the national construction industry are frequent, and there is an urgent need for intelligent and information-based means to achieve remote video monitoring of the entire process of construction projects. At the same time, how to evaluate the quality of the video in real time and with high accuracy is important for remote video accurate deep foundation pits, high slopes, and large formwork displacement monitoring of construction sites, flatness measurement of prefabricated building panels, monitoring of construction lifting machinery and equipment, formwork support systems, and external scaffolding system measurements. It can further improve the supervision efficiency of major hazardous sources on construction sites, improve management efficiency, reduce accidents, and promote the informatization of construction safety production supervision. Therefore, evaluating the distortion of the video can provide guidance for the detection and elimination of video distortion, thereby improving the quality of the video.

[0003] However, today's surveillance video systems are very large, and it is impractical to employ a large number of people to continuously and subjectively evaluate the quality of each surveillance video. Therefore, how to evaluate the quality of surveillance videos in real time and with high accuracy so that the video quality of each subnet can meet the needs of surveillance purposes has become a new direction in the field of surveillance video research. In addition, in the study of objective video quality assessment, human visual characteristics play an important role. Just noticeable difference (JND), as one of the characteristics of human vision, can well reflect the visual redundancy of the human eye and has high application value in quality assessment. However, so far, there are still few JND calculation models that are actually applied to video quality assessment, and most of the JND calculation models have high time complexity and low usability in video quality assessment. Summary of the invention

[0004] The present invention proposes a video quality evaluation method based on just noticeable difference and block effect, which can realize real-time and accurate objective quality evaluation of video.

[0005] The present invention adopts the following technical solutions.

[0006] A video quality evaluation method based on just noticeable difference and blockiness includes the following steps:

[0007] Step S1: Calculate the minimum perceptible difference of pixels according to the change of pixel values ​​between adjacent blocks of the video frame and the visibility threshold of the human visual system for different background brightness values; the perceptible difference is determined by whether the change of the video pixel can be perceived by the human eye as the threshold;

[0008] Step S2: judging the perceptible coding effect of each pixel by the calculated minimum perceptible difference, and applying the judgment result to the detection of the blocking effect;

[0009] Step S3: Detect and extract the salient area of ​​the video using the frame difference method, and use the salient area as a benchmark sample for evaluating the video quality.

[0010] The video frame adjacent blocks are video frame adjacent blocks of a lossless video.

[0011] In the step S1, the minimum noticeable difference of the pixel points is calculated according to the change of the pixel values ​​between adjacent blocks of the video frame; specifically, the following steps are included:

[0012] Step S11: Based on the minimization setting of perceptible difference, the video frame is divided into 2*2 small blocks, and the average difference between the sum of the pixel values ​​of the current block and the sum of the pixel values ​​of the adjacent blocks is calculated; the minimum perceptible difference threshold of the current block is represented as Y jnd , for Y jnd The calculation of is as follows:

[0013] Y jnd =(|A current -A left |+|A current -A right |+|A current -A above |+|A current -A below |) / (4*N1 2 )Formula 1;

[0014] Where N1 represents the size of the block taken, A current , A above , A below , A left , A right Respectively represent the sum of the pixel values ​​of the current block and its corresponding upper, lower, left and right blocks;

[0015] Step S12: Based on the different visibility thresholds of the human visual system for different background brightness values, the parameter F is introduced. jnd (i, j) Correct the calculation result of step S11; the specific calculation is as follows:

[0016]

[0017] Where ref(i,j) represents the pixel value of the video frame at (i,j); F jnd (i, j) means adjustment is made at video frame (i, j), and 0 means no modification is made to the JND threshold;

[0018] Step S13: Add the calculation results of step S11 and step S12 to obtain the minimum perceptible difference threshold in the video frame pixel domain, and use G jnd express;

[0019] Step S14: using the least noticeable difference to judge the distortion of the pixel points of the distorted video frame, the detection model is as follows:

[0020]

[0021] Where dst(i, j) represents the pixel value of the distorted video frame at (i, j); R(i, j) represents the distortion detection result in the pixel domain.

[0022] An R(i,j) value of 1 indicates that the distortion change of the pixel in the pixel domain can be perceived by the human eye, and an R(i,j) value of 0 indicates that the distortion change cannot be perceived by the human eye.

[0023] In step S2, the perceptible coding effect of each pixel is determined by the calculated minimum perceptible difference, and the determination result is applied to the detection of the blocking effect, which specifically includes the following steps:

[0024] Step S21: Process the distorted video frame in blocks, and use the detected distortion of the pixels to detect the blocking effect. The specific calculation is as follows:

[0025]

[0026] Where M represents the size of the block. When the video frame is divided into 8*8 or 16*16 small blocks during compression encoding, M=16 is taken to match the time complexity. block-jnd Indicates the number of distorted pixels in a 16*16 block; block1 indicates the block effect detection result, and a value of 1 indicates the presence of block effect;

[0027] Step S22: Based on the characteristic of block effect that block edges are discontinuous, the vertical edges in each 16*16 small block are detected to see whether the change in pixel value is too great. The specific calculation is as follows:

[0028]

[0029] Among them, block2 indicates the further detection result of the block effect, 1 indicates the existence of block effect, X block-edgeRepresents the sum of the differences between the pixel values ​​of the original video frame and the distorted video frame in one column of a 16*16 small block;

[0030] Step S23: Finally, the determination results of step S21 and step S22 are used as the final blockiness detection result of the video frame. The specific calculation formula is as follows:

[0031] block=block1||block2 Formula 6;

[0032] Where block is the final block effect detection result.

[0033] The step S3 combines the salient region detection algorithm to achieve objective quality evaluation of the video, and specifically includes the following steps:

[0034] Step S31: extracting the salient area of ​​the video based on the frame difference method; converting the read video frame from the three RGB color channels into a single-channel grayscale value image, then subtracting the video frames every preset number of frames to obtain the motion area of ​​the video, and finally using histogram threshold segmentation to obtain a binary image of the salient area;

[0035] Step S32: Based on the calculation result of step S31, the distortion of the video is evaluated in the salient area; the specific calculation is as follows:

[0036] BS=block&Saliency Formula 7;

[0037] Where block is the final result of block effect detection, saliency is the result of saliency area detection, and BS represents the block effect in the saliency area. The specific calculation of the objective quality Q of the video is as follows:

[0038]

[0039] Where W and H represent the width and height of the video frame respectively.

[0040] In step S31, video frame subtraction is performed every 6 frames to obtain the motion region of the video.

[0041] Compared with the prior art, the beneficial effect of the present invention is that it can realize real-time and accurate quality evaluation of videos, and provide guidance for detecting and eliminating video distortion, thereby improving the quality of videos, and has very broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0043] Attached Figure 1 It is a schematic diagram of the workflow of the present invention;

[0044] Attached Figure 2 It is a schematic diagram of the principle of dividing a video frame into blocks in the present invention;

[0045] Attached Figure 3 It is a schematic diagram of the process of obtaining the salient area of ​​the video in the present invention. DETAILED DESCRIPTION

[0046] As shown in the figure, a video quality evaluation method based on the minimum perceptible difference and the block effect includes the following steps:

[0047] Step S1: Calculate the minimum perceptible difference of pixels according to the change of pixel values ​​between adjacent blocks of the video frame and the visibility threshold of the human visual system for different background brightness values; the perceptible difference is determined by whether the change of the video pixel can be perceived by the human eye as the threshold;

[0048] Step S2: judging the perceptible coding effect of each pixel by the calculated minimum perceptible difference, and applying the judgment result to the detection of the blocking effect;

[0049] Step S3: Detect and extract the salient area of ​​the video using the frame difference method, and use the salient area as a benchmark sample for evaluating the video quality.

[0050] The video frame adjacent blocks are video frame adjacent blocks of a lossless video.

[0051] In the step S1, the minimum noticeable difference of the pixel points is calculated according to the change of the pixel values ​​between adjacent blocks of the video frame; specifically, the following steps are included:

[0052] Step S11: Based on the minimization setting of perceptible difference, the video frame is divided into 2*2 small blocks, and the average difference between the sum of the pixel values ​​of the current block and the sum of the pixel values ​​of the adjacent blocks is calculated; the minimum perceptible difference threshold of the current block is represented as Y jnd , for Y jnd The calculation of is as follows:

[0053] Y jnd =(|A current -A left |+|A current -A right |+|A current -A above |+|A current -A below |) / (4*N1 2 )Formula 1;

[0054] Where N1 represents the size of the block taken, A current , A above , A below , Aleft , A right Respectively represent the sum of the pixel values ​​of the current block and its corresponding upper, lower, left and right blocks;

[0055] Step S12: Based on the different visibility thresholds of the human visual system for different background brightness values, the parameter F is introduced. jnd (i, j) Correct the calculation result of step S11; the specific calculation is as follows:

[0056]

[0057] Where ref(i,j) represents the pixel value of the video frame at (i,j); F jnd (i, j) means adjustment is made at video frame (i, j), and 0 means no modification is made to the JND threshold;

[0058] Step S13: Add the calculation results of step S11 and step S12 to obtain the minimum perceptible difference threshold in the video frame pixel domain, and use G jnd express;

[0059] Step S14: using the least noticeable difference to judge the distortion of the pixel points of the distorted video frame, the detection model is as follows:

[0060]

[0061] Where dst(i, j) represents the pixel value of the distorted video frame at (i, j); R(i, j) represents the distortion detection result in the pixel domain.

[0062] An R(i,j) value of 1 indicates that the distortion change of the pixel in the pixel domain can be perceived by the human eye, and an R(i,j) value of 0 indicates that the distortion change cannot be perceived by the human eye.

[0063] In step S2, the perceptible coding effect of each pixel is determined by the calculated minimum perceptible difference, and the determination result is applied to the detection of the blocking effect, which specifically includes the following steps:

[0064] Step S21: Process the distorted video frame in blocks, and use the detected distortion of the pixels to detect the blocking effect. The specific calculation is as follows:

[0065]

[0066] Where M represents the size of the block. When the video frame is divided into 8*8 or 16*16 small blocks during compression encoding, M=16 is taken to match the time complexity. bl ock-j nd represents the number of distorted pixels in a 16*16 block; block1 represents the block effect detection result, and a value of 1 indicates the presence of block effect;

[0067] Step S22: Based on the characteristic of block effect that block edges are discontinuous, the vertical edges in each 16*16 small block are detected to see whether the change in pixel value is too great. The specific calculation is as follows:

[0068]

[0069] Among them, block2 indicates the further detection result of the block effect, 1 indicates the existence of block effect, X block-edge Represents the sum of the differences between the pixel values ​​of the original video frame and the distorted video frame in one column of a 16*16 small block;

[0070] Step S23: Finally, the determination results of step S21 and step S22 are used as the final blockiness detection result of the video frame. The specific calculation formula is as follows:

[0071] block=block1||block2 Formula 6;

[0072] Where block is the final block effect detection result.

[0073] The step S3 combines the salient region detection algorithm to achieve objective quality evaluation of the video, and specifically includes the following steps:

[0074] Step S31: extracting the salient area of ​​the video based on the frame difference method; converting the read video frame from the three RGB color channels into a single-channel grayscale value image, then subtracting the video frames every preset number of frames to obtain the motion area of ​​the video, and finally using histogram threshold segmentation to obtain a binary image of the salient area;

[0075] Step S32: Based on the calculation result of step S31, the distortion of the video is evaluated in the salient area; the specific calculation is as follows:

[0076] BS=block&Saliency Formula 7;

[0077] Where block is the final result of block effect detection, saliency is the result of saliency area detection, and BS represents the block effect in the saliency area. The specific calculation of the objective quality Q of the video is as follows:

[0078]

[0079] Where W and H represent the width and height of the video frame respectively.

[0080] In step S31, video frame subtraction is performed every 6 frames to obtain the motion region of the video.

Claims

1. A video quality assessment method based on just noticeable difference and blockiness, characterized in that: The steps include: Step S1: Calculate the minimum perceptible difference of pixels according to the change of pixel values ​​between adjacent blocks of the video frame and the visibility threshold of the human visual system for different background brightness values; the perceptible difference is determined by whether the change of the video pixel can be perceived by the human eye as the threshold; Step S2: judging the perceptible coding effect of each pixel by the calculated minimum perceptible difference, and applying the judgment result to the detection of the blocking effect; Step S3: Detect and extract the salient area of ​​the video using the frame difference method, and use the salient area as a benchmark sample for evaluating the video quality; The video frame adjacent blocks are video frame adjacent blocks of lossless video; In step S1, the minimum noticeable difference of pixels is calculated according to the change of pixel values ​​between adjacent blocks of the video frame; The specific steps include: Step S11: Based on the minimization setting of perceptible difference, the video frame is divided into 2*2 small blocks, and the average difference between the sum of the pixel values ​​of the current block and the sum of the pixel values ​​of the adjacent blocks is calculated; the minimum perceptible difference threshold of the current block is represented as Y jnd , for Y jnd The calculation of is as follows: Y jnd = (|A current - A left | + |A current - A right | + |A current - A above | + |A current - A below |) / (4 * N1 2 ) Formula 1; Where N1 represents the size of the block taken, A current , A above , A below , A left , A right Respectively represent the sum of the pixel values ​​of the current block and its corresponding upper, lower, left and right blocks; Step S12: Based on the different visibility thresholds of the human visual system for different background brightness values, the parameter F is introduced. jnd (i, j) Correct the calculation result of step S11; the specific calculation is as follows: Where ref(i,j) represents the pixel value of the video frame at (i,j); F jnd (i, j) means adjustment is made at the video frame (i, j), and 0 means no modification is made to the JND threshold; Step S13: Add the calculation results of step S11 and step S12 to obtain the minimum perceptible difference threshold in the video frame pixel domain, and use G jnd express; Step S14: using the least noticeable difference to judge the distortion of the pixel points of the distorted video frame, the detection model is as follows: Where dst(i,j) represents the pixel value of the distorted video frame at (i,j); R(i,j) represents the distortion detection result in the pixel domain; In step S2, the perceptible coding effect of each pixel is determined by the calculated minimum perceptible difference, and the determination result is applied to the detection of the blocking effect, which specifically includes the following steps: Step S21: Process the distorted video frame in blocks, and use the detected distortion of the pixels to detect the blocking effect. The specific calculation is as follows: Where M represents the size of the block. When the video frame is divided into 8*8 or 16*16 small blocks during compression encoding, M=16 is taken to match the time complexity. block-jnd Indicates the number of distorted pixels in a 16*16 block; block1 indicates the block effect detection result, and a value of 1 indicates the presence of block effect; Step S22: Based on the characteristic of block effect that block edges are discontinuous, the vertical edges in each 16*16 small block are detected to see whether the change in pixel value is too great. The specific calculation is as follows: Among them, block2 indicates the further detection result of the block effect, 1 indicates the existence of block effect, X block-edge Represents the sum of the differences between the pixel values ​​of the original video frame and the distorted video frame in one column of a 16*16 small block; Step S23: Finally, the determination results of step S21 and step S22 are used as the final blockiness detection result of the video frame. The specific calculation formula is as follows: block=block1||block2 Formula 6; Where block is the final block effect detection result.

2. The video quality assessment method based on just noticeable difference and blockiness according to claim 1, characterized in that: An R(i,j) value of 1 indicates that the distortion change of the pixel in the pixel domain can be perceived by the human eye, and an R(i,j) value of 0 indicates that the distortion change cannot be perceived by the human eye.

3. The video quality assessment method based on just noticeable difference and blockiness according to claim 1, characterized in that: The step S3 combines the salient region detection algorithm to achieve objective quality evaluation of the video, and specifically includes the following steps: Step S31: extracting the salient area of ​​the video based on the frame difference method; converting the read video frame from the three RGB color channels into a single-channel grayscale value image, then subtracting the video frames every preset number of frames to obtain the motion area of ​​the video, and finally using histogram threshold segmentation to obtain a binary image of the salient area; Step S32: Based on the calculation result of step S31, the distortion of the video is evaluated in the salient area; the specific calculation is as follows: BS=block&Saliency Formula 7; Where block is the final result of block effect detection, saliency is the result of saliency area detection, and BS represents the block effect in the saliency area. The specific calculation of the objective quality Q of the video is as follows: Where W and H represent the width and height of the video frame respectively.

4. The video quality assessment method based on just noticeable difference and blockiness according to claim 3, characterized in that: In step S31, video frame subtraction is performed every 6 frames to obtain the motion region of the video.

Images