CN-122002032-A - Video compression method and system based on just noticeable distortion model of unordered concealment

CN122002032ACN 122002032 ACN122002032 ACN 122002032ACN-122002032-A

Abstract

The invention provides a video compression method and a video compression system based on a just-perceived distortion model of unordered concealment, wherein the method comprises the steps of carrying out improved CSLBP calculation after an input video frame is segmented, weighting and calculating the texture complexity of each image block according to the grouping CSLBP intensity corresponding to each image block, and calculating the texture complexity corresponding to each image block on the premise of not considering quantization distortion according to the texture complexity Value according to And calculating the JND value corresponding to each image block on the premise of considering quantization distortion, and carrying out coding pretreatment on the video frame according to the JND value, so as to reduce gray value redundant information at the corresponding position of the video frame, remove visual redundancy in the video frame and obtain the coded video frame. The invention realizes the technical purpose of imperceptible compression of images, solves the technical problem of high bandwidth requirement in the video image transmission process, and can obtain enough compression gain while improving the coding efficiency.

Inventors

GUO AIYING
YANG PIAO
ZHANG JIANHUA

Assignees

上海大学

Dates

Publication Date: 20260508
Application Date: 20260323

Claims (9)

1. A video compression method based on a just-noticeable distortion model for unordered concealment, comprising: Dividing an input video frame into a plurality of image blocks with preset sizes, then performing improved CSLBP calculation, and weighting and calculating the texture complexity of each image block according to the intensity of the grouping CSLBP corresponding to each image block; According to the texture complexity of each image block, calculating JND quantized value corresponding to each image block under the premise of not considering quantization distortion, and marking as A value; According to The value is calculated as JND quantized value corresponding to each image block under the premise of considering quantization distortion; And carrying out coding pretreatment on the video frame according to the JND value corresponding to each image block, reducing gray value redundant information at the corresponding position of the video frame, and removing visual redundancy in the video frame to obtain the coded video frame.
2. The video compression method based on the just-noticeable distortion model for unordered concealment of claim 1, wherein the improved CSLBP calculation after dividing the input video frame into a plurality of image blocks of a preset size comprises: Dividing the segmented image into a plurality of groups according to the symmetrical relation; according to the brightness masking effect, resetting the weight of the image block with the gray level difference of the corresponding pixel in the group smaller than the average brightness of the center block to 0; And according to CSLBP value calculation formula, weighting calculation is carried out according to the pixel distance in the group to obtain CSLBP value.
3. The video compression method based on the just-noticeable distortion model for unordered concealment according to claim 2, wherein said weighting the texture complexity of each image block according to the packet CSLBP intensity corresponding to each image block comprises: Calculating the unordered intensity of each image block: ; in the formula, For the unordered strength of the nth block, for characterizing the texture complexity of the block, For the CSLBP values of the kth group within the nth block, As the diagonal effect weight of the kth group, For the nth block to be divided into blocks, Is the k-th group in the block; Q is the maximum coding state number of a single group; Wherein: Calculating CSLBP values using a centrosymmetric local binary pattern The calculation mode of (a) is as follows: ; in the formula, For the weighted calculation of the symmetrical pixel gray differences and the center block average gray value, Is the pixel gray value of the symmetrical position, Is the pixel gray value of the symmetric distance, Is a weight value according to the symmetric distance; Measurement For the average brightness masking value of the middle part, the gray level difference and the gray level difference of the pixel points with the same distance from the center block in the corresponding block are used for The comparison is made to obtain a gray scale adjustment weight value, The calculation method is as follows: 。
4. The video compression method based on the just-noticeable distortion model for unordered concealment as claimed in claim 1, wherein the texture complexity of each image block is calculated without considering quantization distortion, based on the texture complexity of each image block Values, including: Model calculation using coefficient reduction And wherein the coefficient reduction-based model is expressed as: ; in the formula, For the unordered strength of the nth partition, the texture complexity of that partition is characterized, For calculation to obtain The value of the sum of the values, Is a constant value, and the constant value is a constant value, In order for the disorder correction factor to be a factor, Is a base threshold deviation; wherein, a disorder degree correction factor is adopted To adjust the threshold of the coefficient reduction-based model, factor the disorder correction The value of (2) is set to Expressed as: ; in the formula, Is the unordered strength of the blocks in the piecewise function.
5. A video compression method based on a just noticeable distortion model for unordered concealment as claimed in claim 1, wherein, according to The value, under the premise of considering quantization distortion, calculates the JND value corresponding to each image block, including: adaptive masking based on background brightness And contrast masking where visibility of an object is reduced in front of another object Respectively performing self-adaptive masking on background brightness Is calculated and contrast masked Wherein: The model for calculating LA masking effect using background luminance is: ; in the formula, For the calculated background luminance adaptive masking, In the horizontal direction of the axis of abscissa, In the vertical coordinate of the drawing, the drawing is, Is a constant value, and is used for the treatment of the skin, Is the background brightness of the pixel point, Is a gray intermediate value; The calculation model for calculating the CM masking effect using visibility is expressed as:) ; ; In the formula, In order to obtain gradient values in four directions by convolution calculation according to convolution kernels g1-g4, In the horizontal direction of the axis of abscissa, In the vertical coordinate of the drawing, the drawing is, For the convolution kernel value, Is the maximum gradient value of four directions; Masking for calculated contrast based on And On the premise of taking quantization distortion into consideration in the calculation model of (2) 、 And Nonlinear combination is carried out to obtain the following JND value calculation model: = ; in the formula, For calculation to obtain The value of the sum of the values, Is a constant.
6. The video compression method based on the just-noticeable distortion model of unordered concealment according to claim 1, wherein the encoding preprocessing is performed on the video frame according to the JND value corresponding to each image block, the gray value redundancy information of the corresponding position of the video frame is reduced, the visual redundancy in the video frame is removed, and the encoded video frame is obtained, comprising: according to the JND value corresponding to each image block, calculating to obtain final human eye imperceptible gray value redundant information; and preprocessing the input image data according to the gray value redundant information, so that the image quality loss perceived by human eyes is not caused while the image transmission bandwidth is reduced.
7. A video compression system based on a just-noticeable distortion model for unordered concealment, comprising: CSLBP a computing module, which is used for performing improved CSLBP computation after dividing an input video frame into a plurality of image blocks with preset sizes, and weighting and computing the texture complexity of each image block according to the intensity of a group CSLBP corresponding to each image block; a JND0 value calculation module for calculating JND quantized value corresponding to each image block according to texture complexity of each image block without considering quantization distortion, and recording as A value; JND value calculation module for calculating JND value according to The value is calculated as JND quantized value corresponding to each image block under the premise of considering quantization distortion; And the video frame coding module is used for carrying out coding pretreatment on the video frame according to the JND value corresponding to each image block, reducing gray value redundant information at the corresponding position of the video frame, and removing visual redundancy in the video frame to obtain the coded video frame.
8. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor is operable to perform the method of any of claims 1-6 when the computer program is executed.
9. A computer readable storage medium having stored thereon a computer program, which, when executed by a processor, is operable to perform the method of any of claims 1-6.

Description

Video compression method and system based on just noticeable distortion model of unordered concealment Technical Field The invention relates to the technical field of digital video compression and visual perception coding, in particular to a video compression method and a video compression system based on a just-noticeable distortion model of unordered concealment, and simultaneously relates to corresponding computer equipment and a computer readable storage medium. Background Today, 8K ultra-high definition, VR/AR, cloud gaming, short video live broadcast are becoming popular, and smartphones, motion cameras, vehicle-mounted visual terminals, and holographic display devices are generating billions of high frame rate, high Dynamic Range (HDR) video streams each day. According to the latest visual network index, video traffic has been more than eight times the total global mobile internet traffic and still grows at about 25% per year, bringing unprecedented bandwidth and storage pressure to 5G/6G air interfaces, edge CDNs, and data centers. Although the compression efficiency of new generation standards such as H.266/VVC, AV1 and AVS3 is improved by 30-50% compared with HEVC, the compression efficiency is still more difficult to approach the shannon limit on the premise of controllable computational complexity by means of the traditional hybrid coding framework (prediction+transformation+entropy coding), and the coding gain curve obviously tends to be flat. In this context, "perceptual video coding" based on the visual characteristics of the human eye becomes a key direction to break through bottlenecks. By introducing a Just Noticeable Distortion (JND) model, the system can further reject visual redundancy on the premise of unchanged subjective quality, and save 15-25% code rate at 1080p/4K gear, thereby remarkably reducing bandwidth requirements and improving play fluency in a weak network environment. Most existing coding methods using a JND model based on DCT transform remove invisible high-frequency information from an input end before coding, so that coding efficiency is improved, but quantization effect is not considered in the process, and obtained compression gain is insufficient. The search finds that: The Chinese patent application No. CN112584153A discloses a video compression method and device based on an just noticeable distortion model, which comprises the steps of carrying out DCT conversion on an image block after inputting a video frame to obtain a spectrogram corresponding to each image block, calculating texture complexity T of each image block, calculating a JND0 value corresponding to each image block on the premise of not considering quantization distortion according to the texture complexity T of each image block, calculating a JND value corresponding to each image block on the premise of considering quantization distortion according to the JND0 value, carrying out coding pretreatment on the video frame according to the JND value corresponding to each image block, reducing the coefficient value of the DCT corresponding to the position of the video frame, and removing visual redundancy in the video frame to obtain a coded video frame. Visual redundancy in the video can be effectively removed, larger compression gain is obtained, and the characteristics of a human eye visual system are more met. However, the following technical problems still exist in the technology of the patent: In the process of calculating the JND value, the complexity of the block can be calculated only after DCT transformation is carried out on the image block of the input video frame. This technique increases the complexity of the computation and requires a high amount of computation. Disclosure of Invention The present invention addresses the above-mentioned deficiencies of the prior art by providing a video compression method and system based on a just noticeable distortion model for unordered concealment, while relating to a corresponding computer device and computer-readable storage medium. According to a first aspect of the present invention, there is provided a video compression method based on a just noticeable distortion model for unordered concealment, comprising: Dividing an input video frame into a plurality of image blocks with preset sizes, then performing improved CSLBP calculation, and weighting and calculating the texture complexity of each image block according to the intensity of the grouping CSLBP corresponding to each image block; According to the texture complexity of each image block, calculating JND quantized value corresponding to each image block under the premise of not considering quantization distortion, and marking as A value; According to The value is calculated as JND quantized value corresponding to each image block under the premise of considering quantization distortion; And carrying out coding pretreatment on the video frame according to the JND value corresponding to each image