WO2021196822A1 - 一种基于自适应自导向滤波的环路滤波方法 - Google Patents

一种基于自适应自导向滤波的环路滤波方法 Download PDF

Info

Publication number
WO2021196822A1
WO2021196822A1 PCT/CN2021/070821 CN2021070821W WO2021196822A1 WO 2021196822 A1 WO2021196822 A1 WO 2021196822A1 CN 2021070821 W CN2021070821 W CN 2021070821W WO 2021196822 A1 WO2021196822 A1 WO 2021196822A1
Authority
WO
WIPO (PCT)
Prior art keywords
filter
filtering
image
ctu
level
Prior art date
Application number
PCT/CN2021/070821
Other languages
English (en)
French (fr)
Inventor
朱策
邓玲玲
蒋妮
王秋月
丁可可
Original Assignee
电子科技大学
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 电子科技大学 filed Critical 电子科技大学
Priority to US17/458,648 priority Critical patent/US11388444B2/en
Publication of WO2021196822A1 publication Critical patent/WO2021196822A1/zh

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/80—Details of filtering operations specially adapted for video compression, e.g. for pixel interpolation
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/80—Details of filtering operations specially adapted for video compression, e.g. for pixel interpolation
    • H04N19/82—Details of filtering operations specially adapted for video compression, e.g. for pixel interpolation involving filtering within a prediction loop
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/169—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding
    • H04N19/17—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being an image region, e.g. an object
    • H04N19/176—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being an image region, e.g. an object the region being a block, e.g. a macroblock
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/70—Denoising; Smoothing
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/102—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
    • H04N19/117—Filters, e.g. for pre-processing or post-processing
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/169—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding
    • H04N19/17—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being an image region, e.g. an object
    • H04N19/172—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being an image region, e.g. an object the region being a picture, frame or field
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/169—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding
    • H04N19/182—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being a pixel
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/169—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding
    • H04N19/186—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being a colour or a chrominance component
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/70—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals characterised by syntax aspects related to video coding, e.g. related to compression standards

Definitions

  • the invention belongs to the technical field of video coding and decoding, and specifically relates to a loop filtering method based on adaptive self-guided filtering.
  • the block-based hybrid coding framework in mainstream video coding standards will cause distortion effects such as block effect, ringing effect, and image blur in the reconstructed image.
  • Loop filtering technology filters the encoded reconstructed image, which can reduce these distortion effects to a certain extent.
  • the impact on the quality of video reconstruction improve the subjective and objective quality of the video.
  • the reconstructed image not only improves the quality of the current reconstructed image, but also provides a reference for the subsequent encoded image. Compared with the image before filtering, the reconstructed image after filtering is more conducive to reference, which can further reduce the prediction residual of subsequent encoding and improve the encoding efficient.
  • a distorted image For a distorted image, it can be regarded as the superposition of the original image and noise, and the purpose of filtering is to remove the noise from the distorted image to obtain the original image. Since natural images are composed of flat areas and texture details, while the filter smooths the noise in the image, it will also smooth the more important texture details in the distorted image. Based on this problem, researchers have proposed multiple edge-preserving filters. , Such as Bilateral Filter (BF), Wiener Filter (WF), and Guided Filter (GF).
  • BF Bilateral Filter
  • WF Wiener Filter
  • GF Guided Filter
  • steering in the steering filter means that in the process of filtering the image p to be filtered to obtain the output image q, it is necessary to use a guiding image I.
  • the guiding image is the image to be filtered itself.
  • An important assumption of the guided filter is that the guided image I and the output image q present a local linear relationship in the filter window, that is, the output at the pixel i can be expressed by the following formula:
  • ⁇ k are the variance and mean value of the guided image pixels in the filter window ⁇ k
  • ⁇ is the regularization coefficient, which is used to punish the larger coefficient a k .
  • a dual self-guided filtering technology based on subspace projection was introduced to further improve coding efficiency.
  • This technology proposes to use two sets of self-guided filters to filter the reconstructed image X to obtain two filtered images X 1 and X 2.
  • the resulting projection X r is closer to the real image Y.
  • X r -X can be expressed linearly with X 1 -X and X 2 -X, so X r can be expressed as:
  • the parameters ⁇ and ⁇ can be calculated according to the least square solution of the matrix equation:
  • the schematic diagram of subspace mapping is shown in Figure 1.
  • the self-guided filter parameters and mapping coefficients are transmitted to the decoding end, and the reconstructed image can be filtered at the decoding end.
  • JVET Compared with HEVC, JVET adds ALF technology to the latest video coding standard VVC under development, which has achieved a greater improvement in coding efficiency, but the various distortion effects in the reconstructed image have not been completely removed, thus further reducing the reconstruction Distortion in the image is still an important requirement in the development of video coding technology.
  • the original dual self-guided filtering technology is not suitable for the VVC standard for high-definition and ultra-high-definition video.
  • the present invention further improves the video quality by improving the dual self-guided filtering technology.
  • the loop filtering method based on the adaptive self-guided filtering of the present invention mainly includes realizing the self-adaptive adjustment of the regularization coefficient ⁇ of the self-guided filter and the iterative optimization of the regional-level mapping.
  • the technical scheme adopted by the present invention is a loop filtering method based on adaptive self-guided filtering.
  • the adaptive self-guided filter can use the local structure information of the reconstructed image to effectively remove the quantization while avoiding the excessive smoothing of the image edges.
  • Noise, while iteratively optimized subspace mapping ensures that the filtering result of subspace mapping is close enough to the original image, including the following steps:
  • the existing technology uses a relatively primitive self-guided filter to filter the entire reconstructed image.
  • the use of this filtering strategy to enhance the quality of the reconstructed image is very inefficient, resulting in the subsequent use of subspace mapping to obtain an unsatisfactory mapping result.
  • the self-guided filter is improved to make the filtering result closer to the original image, then the filtered result obtained by the final mapping will also be closer to the original image, and the filtering effect will be further improved. Since the self-guided filter is only determined by two parameters ⁇ r, ⁇ , the increase of the filter radius r will only make the filtering result smoother, and what is usually needed is to highlight the details of the image texture, so we can consider optimizing the regularization coefficient ⁇ .
  • the ⁇ of the self-guided filter used in the prior art is fixed for all pixels in an image. If a higher ⁇ is used, the texture area will be excessively smooth, while a lower ⁇ will not ensure that the flat area is sufficiently smooth. Therefore, a fixed ⁇ is difficult to balance these two issues. Therefore, the present invention proposes an adaptive self-guided filtering technique that considers the local structure information of the image.
  • the encoder first uses the Sobel operator to calculate the local structure information of the reconstructed image to be filtered, and the 3 ⁇ 3 Sobel operator Convolve with the reconstructed image, calculate the horizontal and vertical pixel gradients, and estimate the edge information of the current reconstructed image:
  • the regularization coefficient ⁇ is adaptively adjusted using G representing the edge information.
  • s is the weight coefficient of the regularization coefficient at the pixel i
  • ⁇ i is the filter window centered on the pixel i
  • ⁇ is a constant with a value of 0.01. If the pixel i is at the strong edge of the image, s is always greater than 1, and if the pixel i is in the flat area of the image, s is always less than 1.
  • the adaptive self-guided filter after adjusting the regularization coefficient ⁇ by using the parameter s has the following characteristics: at strong edges, ⁇ will become smaller, a k is closer to 1, and more texture will be retained; in flat areas , ⁇ will become larger, a k is closer to 0, and b k is closer to the filter window pixel average ⁇ k , making the flat area smoother.
  • the self-adaptive self-guided filter proposed in the present invention can effectively improve the quality of the two filtered images X 1 and X 2 , and is more conducive to subsequent subspace mapping, so that the value of the mapping coefficient will be smaller, which is conducive to reduction
  • the number of bits required for the encoding end to transmit the mapping coefficients improves the encoding efficiency.
  • the dual self-guided filtering technology in the existing AV1 video coding standard uses only a pair of mapping coefficients ⁇ , ⁇ for the entire reconstructed image.
  • the video size is larger and the video content changes more. Large, the mapping result relying only on a pair of mapping coefficients ⁇ , ⁇ is not close enough to the original image.
  • the present invention proposes a regional-level subspace iterative optimization technology.
  • mapping coefficient ⁇ , ⁇ obtained by S2 into the formula (1.4) to obtain the mapping result X r .
  • step S4 Determine whether the CTU level filter flag of the CTU exists in the area and set it from 1 to 0. If yes, recalculate the area level mapping coefficient ⁇ , ⁇ for the CTU with the remaining CTU level filter flag of 1 in the area, and return to step S3, until the filter flag of the CTU in the area no longer changes or the number of iterations reaches 5, ⁇ , ⁇ is determined, otherwise, go to step S5;
  • mapping coefficient ⁇ obtained after iterative optimization, map the CTU with the filter flag of 1 in the region, and compare the rate distortion before and after the iterative optimization using the adaptive self-guided filter and the region-level subspace mapping The cost determines whether to use the filtering operation for the current area.
  • set the area-level filter flag to 1, otherwise set the area-level filter flag to 0.
  • the mapping result of each region is optimal, making the mapping result of the entire image closer to the original image, especially for large-size reconstructed images, further improving the quality of reconstructed images , It also brings a lot of additional filtering information, which requires an effective way of encoding.
  • the optimal rate-distortion cost RDCost1 obtained by traversing the 32 sets of filter parameters in the self-guided filter parameter table for adaptive self-guided filtering and regional subspace iterative optimization is still higher than the rate-distortion cost RDCost0 when the filter is not used, then
  • the image-level filtering flag of the reconstructed image is set to 0, otherwise the image-level filtering flag is set to 1, and the relevant information of the filter is transmitted to the decoding end.
  • mapping coefficient ⁇ , ⁇ of each region is basically distributed in [-5, +5]. Therefore, for the integer part of the coefficient, the 0-order exponential Golomb coding is used directly; for the decimal part of the coefficient, the 6-bit precision is used, that is, the decimal part is shifted to the left by 6 bits and then rounded, and the 0-order exponential Golomb coding is used, so code a
  • the mapping coefficient requires up to 9bit codewords.
  • the specific method of the self-guided filtering is to adaptively adjust the decisive parameter regularization coefficients of the self-guided filter to realize the adaptive self-guided filter based on the local structure information of the reconstructed image, and filter the adaptive self-guided filter As a result, iterative optimization of subspace mapping is used to further reduce the gap between the filtered image and the original image and improve the quality of the reconstructed image.
  • the decoding end When the decoding end reconstructs the coded image, it first decodes the image-level filter flag of the current reconstructed image. If the image-level filter flag is 0, it means that the current reconstructed image does not need to be filtered, and no operation is performed on the reconstructed image at this time; if If the frame-level filter flag is 1, you need to continue to decode the self-guided filter parameter index, find the filter parameter corresponding to the decoded index in the self-guided filter parameter table, use the filter parameter to perform adaptive self-guided filtering, and get two filters The result is X 1 , X 2 .
  • the reconstructed image Like the encoding end, divide the reconstructed image into multiple regions of the same size, and continue to decode the region-level filter flags and region-level mapping coefficients of each region. For regions with a region-level filter flag of 1, further decoding of the CTU-level filter flag is required Bit, if there is a CTU with a filter flag of 0, the reconstructed pixel of the CTU uses the reconstructed pixel before filtering, if the filter flag of the CTU is 1, then the two sets of adaptive self-guided filtering results of the CTU use the corresponding regional level The mapping coefficient implements subspace mapping, and the final reconstructed pixel of the CTU is replaced with the reconstructed pixel after filtering.
  • the multiplier is determined by the encoder.
  • the beneficial effects of the present invention are that the present invention can improve the filtering effect of the loop filter, reduce the distortion effect of the reconstructed image, and improve the reconstruction quality of the video frame, so that the reconstructed image is more conducive to subsequent reference, and further improves the subjective and objective quality of the video.
  • Figure 1 shows the operable rate-distortion curve
  • Figure 2 is a schematic diagram of the structure of the time-domain distortion propagation chain under the LD structure
  • Figure 3 is a schematic diagram of the LD encoding structure
  • Figure 4 shows the rate-distortion curve of the Fourpeople sequence.
  • VTM3.0 As the experimental platform, add the dual self-guided filtering algorithm based on VVC optimization proposed in this chapter between SAO and ALF in the loop filter module, and verify it under the configuration of AI, LD and RA.
  • the test condition is VTM Common Test Conditions (CTC), including 22 sequences of Class A to Class F.
  • CTC Common Test Conditions
  • the test results show the coding performance of each test sequence, the average performance of each Class, and the average performance of all test sequences.
  • the quantization parameter QP is 22, 27, 32, 37.
  • BD-Rate is used as the comparison standard to indicate the change of bit rate under the same reconstruction quality.
  • the BD-Rate is a negative value, it means that the code rate is reduced under the same reconstruction quality and the technology has a gain.
  • the BD-Rate is positive, it means that the code rate is increased under the same reconstruction quality and the technology has a loss.
  • Table 1 shows the BD-Rate of the luminance Y component of the present invention in the three coding configurations of AI, LB and RA compared with VTM3.0. It can be clearly seen that the present invention is in the coding configuration of AI, LB and RA. Its coding performance has been improved.
  • the luminance component has an average code rate saving of 0.23%; under the LDB coding structure, the luminance component has an average code rate saving of 0.16%; under the RA coding structure, The luminance component has an average code rate saving of 0.23%.
  • Table 1 The code rate saving of the present invention compared to VTM3.0
  • the rate-distortion curve of the BQSquare sequence shown in Figure 4 the abscissa represents the bit rate required to encode the video sequence, the ordinate represents the PSNR of the Y component of the compressed video sequence, and the blue line represents the introduction of the loop proposed in this section.
  • the rate-distortion curve of the encoded BQSquare sequence is encoded, and the shape mark curve represents the rate-distortion curve of the encoded BQSquare sequence of the original VTM3.0. It can be seen from Figure 4 that this algorithm brings a certain coding improvement in the luminance component.
  • the coding complexity of this algorithm increases by 7%, 4%, and 4% under the coding structure of AI, RA, and LDB respectively.
  • the coding time increases more. This is because the self-directed filter needs to be Filtering is performed on each reconstructed pixel, and a larger video size has more reconstructed pixels, resulting in an increase in time overhead.

Landscapes

  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Compression Or Coding Systems Of Tv Signals (AREA)
  • Image Processing (AREA)

Abstract

本发明属于视频编解码技术领域，具体涉及一种基于自适应自导向滤波的环路滤波方法。本发明的方法主要是考虑图像的局部结构信息的自适应自导向滤波技术，通过计算正则化系数的权重系数，对正则化系数进行调整，实现自适应自导向滤波，然后对每个区域实施的区域级子空间映射迭代优化，使每个区域的映射结果达到最优，使得整幅图像的映射结果更加接近原始图像。本发明的有益效果为，能提高环路滤波的滤波效果，减少重建图像的失真效应，提高视频帧的重建质量，使得重建图像更有利于后续参考，进一步提高视频的主客观质量。

Description

一种基于自适应自导向滤波的环路滤波方法 技术领域
本发明属于视频编解码技术领域，具体涉及一种基于自适应自导向滤波的环路滤波方法。
背景技术
主流视频编码标准中基于块的混合编码框架会导致重建图像中存在方块效应、振铃效应以及图像模糊等失真效应，环路滤波技术对已编码重建图像进行滤波，一定程度上可以降低这些失真效应对视频重建质量的影响，提升视频的主客观质量。重建图像在滤波后不仅提升当前重建图像的质量，还为后续编码图像提供参考，相比滤波前的图像，滤波后的重建图像更有利于参考，可以进一步缩小后续编码的预测残差，提高编码效率。
对于失真图像，可以将其看成原始图像与噪声的叠加，而滤波的目的是将噪声从失真图像中剔除，从而得到原始图像。由于自然图像是由平缓区域和纹理细节共同组成，在滤波器平滑图像中的噪声的同时，会将失真图像中较为重要的纹理细节一并平滑，基于该问题研究人员提出多个保边滤波器，如双边滤波器(Bilateral Filter，BF)、维纳滤波器(Wiener Filter，WF)，以及导向滤波器(Guided Filter，GF)等。
导向滤波器中的导向一词是指对待滤波图像p进行滤波得到输出图像q的过程中,需要利用一个引导图像I，对于自导向滤波器，引导图像即为待滤波图像本身。导向滤波器的一个重要假设是引导图像I与输出图像q在滤波窗口呈现局部线性关系,即在像素i处的输出可用下式表示：
q i＝a kI i+b k,
Figure PCTCN2021070821-appb-000001
由于待滤波图像p可视作输出图像q受噪声污染产生的退化图像，假设噪声为n，则存在q i＝p i-n i。
通过求解min n 2,可以计算得到滤波窗口(2r+1)×(2r+1)内的局部线性模型的系数a k和截距b k:
Figure PCTCN2021070821-appb-000002
b k＝(1-a k)μ k          (0.3)
Figure PCTCN2021070821-appb-000003
和μ k是滤波窗口ω k内引导图像像素的方差和均值,ε是正则化系数，用于惩罚较大的系数a k。
当ε＝0，a k＝1并且b k＝0，此时输出图像像素等于引导图像像素；当ε>0时，有以下两种情况。
情况1：
Figure PCTCN2021070821-appb-000004
较大的区域，图片在滤波窗口ω k中变化剧烈。此时有
Figure PCTCN2021070821-appb-000005
故a k≈1，b k≈0，q≈p，图像边缘被保留。
情况2：
Figure PCTCN2021070821-appb-000006
较小的区域，图片在滤波窗口ω k中几乎不变。此时有
Figure PCTCN2021070821-appb-000007
故a k≈1，b k≈μ k，q≈μ k，图像区域被平滑。
AV1视频编码标准制定过程中引入了基于子空间投影的双自导向滤波技术，进一步提升编码效率。该技术提出使用两组自导向滤波器对重建图像X滤波，得到两个滤波图像X 1和X 2，通过将真实图像Y与重建图像X的差值矢量投影到X 1与X的差值矢量(X 1-X)和X 2与X的差值矢量(X 2-X)所生成的子空间，所得到的投影X r与真实图像Y更接近。X r-X可用X 1-X和X 2-X线性表出，故X r可表示为：
X r＝X+α(X 1-X)+β(X 2-X)        (0.4)
当编码器给出原始图像Y、待滤波重建图像X以及滤波图像X 1、X 2，便可以根据矩阵方程的最小二乘解计算出参数α、β：
{α,β} T＝(A TA) -1A Tb         (0.5)
其中A＝{X 1-X,X 2-X}，b＝Y-X，公式(1.5)中的图像矩阵均表示为列向量。
子空间映射示意图如图1所示。将自导向滤波器参数以及映射系数传输至解码端，便可在解码端对重建图像进行滤波。
相较于HEVC，JVET向正在研制的最新视频编码标准VVC中添加了ALF技术，获取了较大的编码效率的提升，但重建图像中的各种失真效应依然没有被完全去除，因而进一步降低重建图像中的失真依然是视频编码技术发展过程中的重要需求。同时，原始的双自导向滤波技术并不适用于针对高清、超高清视频的VVC标准。
发明内容
针对上述问题，本发明通过对双自导向滤波技术进行改进，进一步提升视频质量。本发明基于自适应自导向滤波的环路滤波方法主要包括实现自导向滤波器的正则化系数ε自适应调整以及区域级映射迭代优化。
本发明采用的技术方案是，一种基于自适应自导向滤波的环路滤波方法，自适应的自导向滤波器能利用重建图像的局部结构信息在避免图像边缘被过度平滑的同时有效地去除量化噪声，同时迭代优化的子空间映射保证了子空间映射的滤波结果足够接近原始图像，包括以下步骤：
一、对重建图像实施自适应自导向滤波操作:
现有的技术对整幅重建图像使用较为原始的自导向滤波器进行滤波，采用这种滤波策略增强重建图像质量非常低效，导致后续使用子空间映射得到的映射结果也不理想，如果可以对自导向滤波器进行改进，使滤波结果更加接近原始图像，那么最终映射得到的滤波结果也将更接近原始图像，滤波效果也会进一步提升。由于自导向滤波器仅决定于两个参数{r,ε}，滤波半径r增大只会使得滤波结果更加平滑，而通常需要的是突出图像纹理细节，因此可以考虑优化正则化系数ε。现有技术使用的自导向滤波器的ε对于一张图像中的所有像素是固定不变的，若采用较高的ε会导致纹理区域过度平滑，而较低的ε难以保证平坦区域足够平滑，因此固定的ε难以平衡这两个问题。因此本发明提出考虑图像的局部结构信息的自适应自导向滤波技术。
在编码端和解码端设置自导向滤波参数表存放32组两两一组的自导向滤波参数{r 1,ε 1,r 2,ε 2}，将图像级滤波标志初始化为1，使用每一组自导向滤波参数对重建图像进行滤波得到两个滤波结果X 1、X 2，在滤波时编码器首先利用Sobel算子计算当前待滤波重建图像 的局部结构信息，将3×3的Sobel算子与重建图像卷积，分别计算出横向和纵向的像素梯度，估计当前重建图像的边缘信息：
Figure PCTCN2021070821-appb-000008
将横向梯度和纵向梯度结合：
Figure PCTCN2021070821-appb-000009
对于滤波参数表中的一组滤波参数{r 1,ε 1}、{r 2,ε 2}，使用表征边缘信息的G对正则化系数ε自适应调整。
Figure PCTCN2021070821-appb-000010
Figure PCTCN2021070821-appb-000011
s是像素i处的正则化系数的权重系数，ω i是以像素i为中心的滤波窗口，γ是取值为0.01的常数。如果像素i处于图像的强边缘处，则s总是大于1，若像素i处于图像的平坦区域，则s总是小于1。通过使用参数s对正则化系数ε进行调整后的自适应自导向滤波器具有以下特点：在强边缘处，ε会变小，a k更接近1，更多的纹理会被保留；在平坦区域，ε会变大，a k更接近0，b k更接近滤波窗口像素均值μ k，使得平坦区域更加平滑。
本发明提出的自适应自导向滤波器对重建图像进行滤波能有效提高两个滤波图像X 1、X 2的质量，更有利于后续子空间映射，使得映射系数的值会变小，有利于降低编码端传输映射系数时所需比特数，提高编码效率。
二、区域级子空间映射迭代优化：
现有的AV1视频编码标准中的双自导向滤波技术对整幅重建图像仅使用一对映射系数 {α,β}，然而对于高分辨率视频例如4K视频，视频尺寸较大，视频内容变化较大，仅依靠一对映射系数{α,β}的映射结果并不足够接近原始图像。如果可以将一幅较大尺寸的图像分割成多个不重叠的区域，对每个区域采用不同的映射系数，并且使用区域级的滤波标志记录是否对当前区域进行滤波，将使得各个区域的映射效果更好，从而提高整幅图片的滤波效率，因此本发明提出区域级子空间迭代优化技术。
对每个区域实施的区域级子空间映射迭代优化步骤如下：
S1、将图像分割为多个大小相等互不重叠的区域，区域的边界与CTU的边界对齐，将区域级滤波标志均初始化为1；
S2、将每个区域中CTU的CTU级滤波标志初始化为1，并根据公式(1.5)计算每个区域的映射系数。
S3、将S2得到的映射系数{α,β}代入公式(1.4)得到映射结果X r。仅对区域中的每个滤波标志为1的CTU进行子空间映射操作，根据RDO对当前区域中的CTU逐个决定是否使用滤波器，若决定不使用滤波器，则将该CTU的CTU级滤波标志设置为0；
S4、判断区域中是否存在CTU的CTU级滤波标志从1置为0，若是，则对该区域剩余CTU级滤波标志为1的CTU重新计算区域级映射系数{α,β}，并回到步骤S3，直至该区域中CTU的滤波器标志不再改变或迭代次数达到5次，{α,β}确定，否则进入步骤S5；
S5、根据迭代优化后得到的最终的映射系数{α,β}对区域中滤波标志为1的CTU进行映射，通过比较使用自适应自导向滤波器以及区域级子空间映射迭代优化前后的率失真代价决定是否对当前区域使用滤波操作，对当前区域进行滤波时设置区域级滤波标志为1，否则设置区域级滤波标志为0。
在对每个区域实施区域级子空间映射迭代优化后，每个区域的映射结果达到最优，使得整幅图像的映射结果更加接近原始图像，特别是针对大尺寸重建图像，进一步提升重建图像质量，也因此带来大量附加的滤波信息，需要有效的方式进行编码。
三、滤波附加信息熵编码：
若遍历自导向滤波器参数表中的32组滤波参数进行自适应自导向滤波和区域级子空间 迭代优化得到的最优率失真代价RDCost1依然高于不使用滤波器时的率失真代价RDCost0，则将该重建图像的图像级滤波标志设置为0，否则图像级滤波标志设置为1，并传输滤波器的相关信息至解码端。
判断图像级滤波标志是否为0，若是，则传输一个为0的码字至解码端，告知解码端不需对当前重建图像实施滤波操作。否则，传输一个为1的码字至解码端，同时传输滤波器的相关信息：包括根据RDO遍历滤波参数表得到的最优滤波参数索引、每个区域的区域级滤波标志、区域内的每个CTU的CTU级滤波标志以及所有区域的区域级映射系数{α,β}。
每个区域的映射系数{α,β}基本分布于[-5,+5]。故对于系数的整数部分，直接采用0阶指数哥伦布编码；对于系数的小数部分，使用6比特精度表示，即对小数部分左移6位后取整，并使用0阶指数哥伦布编码，因此编码一个映射系数最多需要9bit码字。
所述自导向滤波的具体方法是对自导向滤波器的决定性参数正则化系数进行自适应调整，实现基于重建图像的局部结构信息的自适应自导向滤波器，对自适应自导向滤波器的滤波结果采用子空间映射迭代优化进一步减小滤波图像与原始图像之间的差距，提高重建图像质量。
解码端对编码图像进行重建时，先解码出当前重建图像的图像级滤波标志位，若图像级滤波标志为0，则表示当前重建图像不需要进行滤波，此时不对重建图像做任何操作；若帧级滤波标志为1，则还需要继续解码自导向滤波参数索引，在自导向滤波参数表中找到解码出的索引对应的滤波参数，使用该滤波参数进行自适应自导向滤波，得到两个滤波结果X 1、X 2。与编码端一样，将重建图像划分为多个大小一致的区域，继续解码每个区域的区域级滤波标志位以及区域级映射系数，对于区域级滤波标志为1的区域需进一步解码CTU级滤波标志位，若存在CTU的滤波标志为0，则该CTU的重建像素使用滤波前的重建像素，若CTU的滤波标志为1，则对该CTU的两组自适应自导向滤波结果使用对应的区域级映射系数实现子空间映射，将该CTU最终的重建像素使用滤波后的重建像素替代。
率失真代价计算公式为RDCost＝D+λR，其中D表示当前编码单元的失真，R表示编码对当前编码单元实施本发明技术中的滤波操作所需信息消耗的比特数，λ是拉格朗日乘子，由编码器决定。
本发明的有益效果为，本发明能提高环路滤波的滤波效果，减少重建图像的失真效应， 提高视频帧的重建质量，使得重建图像更有利于后续参考，进一步提高视频的主客观质量。
附图说明
图1为可操作率失真曲线
图2为LD结构下时域失真传播链的构造示意图
图3为LD编码结构示意图
图4为Fourpeople序列的率失真曲线图。
具体实施方式
下面使用仿真示例说明本发明方案的有效性：
以VTM3.0为实验平台，在环路滤波模块中的SAO和ALF之间加入本章提出的基于VVC优化的双自导向滤波算法，并在AI、LD和RA配置下进行验证，测试条件为VTM的通用测试条件(Common Test Conditions，CTC)，包括Class A～Class F共22个序列，测试结果展示每个测试序列的编码性能、每个Class的性能均值以及所有测试序列的性能均值，测试的量化参数QP为22，27，32，37。
将实验结果与原始的参考软件VTM3.0对比，使用BD-Rate作为比较标准，表示在相同重建质量下码率的变化。当BD-Rate为负值时，代表相同重建质量下码率减少，技术具有增益，当BD-Rate为正值时，代表相同重建质量下码率增加，技术存在损失。表1给出了相比与VTM3.0，本发明分别在AI、LB和RA三种编码配置下的亮度Y分量的BD-Rate，可明显看出本发明在AI、LB和RA编码配置下其编码性能均取得了提升，在AI编码结构下，亮度分量平均有0.23％的码率节省；在LDB编码结构下，对于亮度分量平均有0.16％的码率节省；在RA编码结构下，对于亮度分量平均有0.23％的码率节省。
表1 相比于VTM3.0本发明的码率节省
Figure PCTCN2021070821-appb-000012
Figure PCTCN2021070821-appb-000013
Figure PCTCN2021070821-appb-000014
如图4所示的BQSquare序列的率失真曲线，横坐标表示编码该视频序列所需码率，纵坐标表示压缩后的视频序列Y分量的PSNR，蓝色的线表示引入本节提出的环路滤波算法后编码BQSquare序列的率失真曲线，放形标记曲线表示原始VTM3.0的编码BQSquare序列的率失真曲线。由图4可看出该算法在亮度分量带来了一定的编码提升。
在编码复杂度方面，使用编码时间增加百分比衡量：
Figure PCTCN2021070821-appb-000015
表2 相比于VTM3.0本发明的编码时间增加百分比
Figure PCTCN2021070821-appb-000016
该算法在AI、RA、LDB编码结构下编码复杂度分别增加7％、4％、4％，并且随着视频尺寸的增加，编码时间增加的越多，这是由于自导向滤波器需要对每个重建像素进行滤波，较大的视频尺寸有更多的重建像素，导致时间开销增大。

Claims (1)

  1. 一种基于自适应自导向滤波的环路滤波方法，其特征在于，包括以下步骤：
    S1、对重建图像进行自适应自导向滤波，具体为：
    在编码端和解码端设置自导向滤波参数表存放32组两两一组的自导向滤波参数{r 1,ε 1,r 2,ε 2}，将图像级滤波标志初始化为1，使用每一组自导向滤波参数对重建图像X进行自适应自导向滤波得到两个滤波结果X 1、X 2，在滤波时编码器首先利用Sobel算子计算当前待滤波重建图像的局部结构信息，将3×3的Sobel算子与重建图像卷积，分别计算出横向和纵向的像素梯度：
    Figure PCTCN2021070821-appb-100001
    将横向梯度和纵向梯度结合，估计当前重建图像的边缘信息：
    Figure PCTCN2021070821-appb-100002
    对于滤波参数表中的一组滤波参数{r 1,ε 1}、{r 2,ε 2}，使用表征边缘信息的G对正则化系数ε自适应调整：
    Figure PCTCN2021070821-appb-100003
    Figure PCTCN2021070821-appb-100004
    s是像素i处的正则化系数的权重系数，ω i是以像素i为中心的滤波窗口，γ是取值为0.01的常数；如果像素i处于图像的强边缘处，则s总是大于1，若像素i处于图像的平坦区域，则s总是小于1，通过使用参数s对正则化系数ε进行调整后的自适应自导向滤波器具有以下 特点：在强边缘处，ε会变小，a k更接近1，更多的纹理会被保留；在平坦区域，ε会变大，a k更接近0，b k＝(1-a k)μ k，即b k更接近滤波窗口像素均值μ k，使得平坦区域更加平滑；a k是滤波窗口内的局部线性模型的系数，b k是滤波窗口内的局部线性模型的截距，在滤波窗口内的待滤波像素与输出像素存在以下局部线性模型，根据该局部线性模型即可求得像素i处的输出像素q i：
    Figure PCTCN2021070821-appb-100005
    I i为引导图像；
    S2、对每个区域实施区域级子空间映射迭代，具体包括：
    S21、将图像分割为多个大小相等互不重叠的区域，区域的边界与CTU的边界对齐，将区域级滤波标志均初始化为1；
    S22、将每个区域中CTU的CTU级滤波标志初始化为1，并根据如下公式计算每个区域的映射系数：
    {α,β} T＝(A TA) -1A Tb
    其中A＝{X 1-X,X 2-X}，b＝Y-X，公式中的图像矩阵均为列向量，Y为原始图像；
    S23、将S22得到的映射系数{α,β}代入公式
    X r＝X+α(X 1-X)+β(X 2-X)
    得到映射结果X r，仅对区域中的每个滤波标志为1的CTU进行子空间映射操作，根据RDO对当前区域中的CTU逐个决定是否使用滤波器，若决定不使用滤波器，则将该CTU的CTU级滤波标志设置为0；
    S24、判断区域中是否存在CTU的CTU级滤波标志从1置为0，若是，则对该区域剩余CTU级滤波标志为1的CTU重新计算区域级映射系数{α,β}，并回到步骤S23，直至该区域中CTU的滤波器标志不再改变或迭代次数达到5次，{α,β}确定，否则进入步骤S25；
    S25、根据迭代优化后得到的最终的映射系数{α,β}对区域中滤波标志为1的CTU进行映射，通过比较使用自适应自导向滤波器以及区域级子空间映射迭代优化前后的率失真代价决定是否对当前区域使用滤波操作，对当前区域进行滤波时设置区域级滤波标志为1，否则设置区域级滤波标志为0；率失真代价计算公式为RDCost＝D+λR，其中D表示当前编码单元的失真，R表示编码对当前编码单元实施滤波操作所需信息消耗的比特数，λ是拉格朗日乘子，由编码器决定；
    S3、滤波附加信息熵编码：
    S31、若遍历自导向滤波器参数表中的32组滤波参数进行自适应自导向滤波和区域级子空间映射优化得到的最优率失真代价RDCost1高于不使用滤波器时的率失真代价RDCost0，则将该重建图像的图像级滤波标志设置为0，否则图像级滤波标志设置为1；
    S32、判断图像级滤波标志是否为0，若是，则传输一个为0的码字至解码端，告知解码端不需对当前重建图像实施滤波操作；否则，传输一个为1的码字至解码端，同时传输滤波器的信息，包括：根据RDO遍历滤波参数表得到的最优滤波参数索引、每个区域的区域级滤波标志、区域内的每个CTU的CTU级滤波标志以及所有区域的区域级映射系数{α,β}；
    S4、解码端对编码图像进行重建时，先解码出当前重建图像的图像级滤波标志位，若图像级滤波标志为0，则表示当前重建图像不需要进行滤波，此时不对重建图像做任何操作；若图像级滤波标志为1，则还需要继续解码自导向滤波参数索引，在自导向滤波参数表中找到解码出的索引对应的滤波参数，使用该滤波参数进行自适应自导向滤波，得到两个滤波结 果X 1、X 2；与编码端一样，将重建图像划分为多个大小一致的区域，继续解码每个区域的区域级滤波标志位以及区域级映射系数，对于区域级滤波标志为1的区域需进一步解码CTU级滤波标志位，若存在CTU的滤波标志为0，则该CTU的重建像素使用滤波前的重建像素，若CTU的滤波标志为1，则对该CTU的两组自适应自导向滤波结果使用对应的区域级映射系数实现子空间映射，将该CTU最终的重建像素使用滤波后的重建像素替代。
PCT/CN2021/070821 2020-03-31 2021-01-08 一种基于自适应自导向滤波的环路滤波方法 WO2021196822A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US17/458,648 US11388444B2 (en) 2020-03-31 2021-08-27 Loop filtering method based on adaptive self-guided filtering

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010242336.4 2020-03-31
CN202010242336.4A CN111314711B (zh) 2020-03-31 2020-03-31 一种基于自适应自导向滤波的环路滤波方法

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US17/458,648 Continuation US11388444B2 (en) 2020-03-31 2021-08-27 Loop filtering method based on adaptive self-guided filtering

Publications (1)

Publication Number Publication Date
WO2021196822A1 true WO2021196822A1 (zh) 2021-10-07

Family

ID=71147493

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/070821 WO2021196822A1 (zh) 2020-03-31 2021-01-08 一种基于自适应自导向滤波的环路滤波方法

Country Status (3)

Country Link
US (1) US11388444B2 (zh)
CN (1) CN111314711B (zh)
WO (1) WO2021196822A1 (zh)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113949872A (zh) * 2021-11-09 2022-01-18 华侨大学 一种基于3D-Gradient引导的屏幕内容视频编码码率控制方法
CN116452594A (zh) * 2023-06-19 2023-07-18 安徽百胜电子系统集成有限责任公司 一种输电线路状态可视化监测预警方法及系统
CN116525073A (zh) * 2023-07-03 2023-08-01 山东第一医科大学第一附属医院(山东省千佛山医院) 一种基于健康体检大数据的数据库智能管理系统
CN116740058A (zh) * 2023-08-11 2023-09-12 深圳市金胜电子科技有限公司 一种固态硬盘配套晶圆的质量检测方法

Families Citing this family (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111314711B (zh) * 2020-03-31 2021-05-14 电子科技大学 一种基于自适应自导向滤波的环路滤波方法
CN111698503B (zh) * 2020-06-22 2022-09-09 深圳市迪威码半导体有限公司 一种基于预处理的视频高倍压缩方法
CN111735448A (zh) * 2020-06-23 2020-10-02 上海航天控制技术研究所 一种星图联合非均匀校正方法、设备及存储介质
WO2022218285A1 (en) * 2021-04-12 2022-10-20 Beijing Bytedance Network Technology Co., Ltd. Guided filter usage
CN113365060B (zh) * 2021-05-21 2022-04-26 北京奇艺世纪科技有限公司 一种图像滤波方法、装置、设备及存储介质
CN114125471A (zh) * 2021-11-27 2022-03-01 北京工业大学 一种视频编码前置滤波方法
WO2023231008A1 (zh) * 2022-06-02 2023-12-07 Oppo广东移动通信有限公司 编解码方法、编码器、解码器以及存储介质
CN116188305B (zh) * 2023-02-16 2023-12-19 长春理工大学 基于加权引导滤波的多光谱图像重建方法
CN117237238B (zh) * 2023-11-13 2024-03-29 孔像汽车科技(上海)有限公司 基于导向滤波的图像去噪方法、系统、设备及存储介质

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101783939A (zh) * 2009-01-16 2010-07-21 复旦大学 一种基于人眼视觉特性的图像编码方法
CN103051890A (zh) * 2011-09-27 2013-04-17 美国博通公司 根据视频编码进行自适应环路滤波
CN104735450A (zh) * 2015-02-26 2015-06-24 北京大学 一种在视频编解码中进行自适应环路滤波的方法及装置
CN105791877A (zh) * 2016-03-15 2016-07-20 北京大学 视频编解码中自适应环路滤波的方法
CN111314711A (zh) * 2020-03-31 2020-06-19 电子科技大学 一种基于自适应自导向滤波的环路滤波方法

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8402098B2 (en) * 2009-08-13 2013-03-19 Clark C. Dircz System and method for intelligence gathering and analysis
CN106878729B (zh) * 2010-10-05 2019-09-24 寰发股份有限公司 基于分区基础的自适应环路滤波方法和装置

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101783939A (zh) * 2009-01-16 2010-07-21 复旦大学 一种基于人眼视觉特性的图像编码方法
CN103051890A (zh) * 2011-09-27 2013-04-17 美国博通公司 根据视频编码进行自适应环路滤波
CN104735450A (zh) * 2015-02-26 2015-06-24 北京大学 一种在视频编解码中进行自适应环路滤波的方法及装置
CN105791877A (zh) * 2016-03-15 2016-07-20 北京大学 视频编解码中自适应环路滤波的方法
CN111314711A (zh) * 2020-03-31 2020-06-19 电子科技大学 一种基于自适应自导向滤波的环路滤波方法

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
LINGLING DENG: "Research on the Loop Filter and Rate Distortion Optimization of Versatile Video Coding", CHINESE MASTER'S THESES FULL-TEXT DATABASE, no. 07, 1 March 2020 (2020-03-01), pages 1 - 81, XP055855828, DOI: 10.27005/d.cnki.gdzku.2020.003498 *

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113949872A (zh) * 2021-11-09 2022-01-18 华侨大学 一种基于3D-Gradient引导的屏幕内容视频编码码率控制方法
CN116452594A (zh) * 2023-06-19 2023-07-18 安徽百胜电子系统集成有限责任公司 一种输电线路状态可视化监测预警方法及系统
CN116452594B (zh) * 2023-06-19 2023-08-29 安徽百胜电子系统集成有限责任公司 一种输电线路状态可视化监测预警方法及系统
CN116525073A (zh) * 2023-07-03 2023-08-01 山东第一医科大学第一附属医院(山东省千佛山医院) 一种基于健康体检大数据的数据库智能管理系统
CN116525073B (zh) * 2023-07-03 2023-09-15 山东第一医科大学第一附属医院(山东省千佛山医院) 一种基于健康体检大数据的数据库智能管理系统
CN116740058A (zh) * 2023-08-11 2023-09-12 深圳市金胜电子科技有限公司 一种固态硬盘配套晶圆的质量检测方法
CN116740058B (zh) * 2023-08-11 2023-12-01 深圳市金胜电子科技有限公司 一种固态硬盘配套晶圆的质量检测方法

Also Published As

Publication number Publication date
CN111314711B (zh) 2021-05-14
US20210392382A1 (en) 2021-12-16
US11388444B2 (en) 2022-07-12
CN111314711A (zh) 2020-06-19

Similar Documents

Publication Publication Date Title
WO2021196822A1 (zh) 一种基于自适应自导向滤波的环路滤波方法
TWI787460B (zh) 利用率失真最佳化在視訊編碼之影像再成形
CN108184129B (zh) 一种视频编解码方法、装置及用于图像滤波的神经网络
JP2021044811A (ja) サンプルアダプティブオフセット制御
WO2022027881A1 (zh) 基于视频序列特征和QP-λ修正的时域率失真优化方法
TWI702827B (zh) 視訊編解碼方法/裝置和相應地非易失性計算機可讀介質
KR102350436B1 (ko) 영역 기반 필터에 대해서 협력적 분할 부호화하는 방법 및 장치
US10728549B2 (en) Adaptive loop filtering for high-dynamic range video
CN1301370A (zh) 用于减少压缩的视频中的喘息污迹的方法和设备
WO2020177133A1 (zh) 环路滤波实现方法、装置及计算机存储介质
WO2021120614A1 (zh) 二次编码优化方法
WO2018095890A1 (en) Methods and apparatuses for encoding and decoding video based on perceptual metric classification
CN111131819A (zh) 依赖性量化的编码工具下的量化参数
US20230209051A1 (en) Filtering method and apparatus, and device
Wang et al. An integrated CNN-based post processing filter for intra frame in versatile video coding
CN107295353B (zh) 视频编码参考块滤波方法
TWI780742B (zh) 編解碼方法及裝置
WO2020140219A1 (zh) 帧内预测方法、装置及计算机存储介质
CN115988201B (zh) 一种编码胶片颗粒的方法、装置、电子设备和存储介质
CN113132736B (zh) 一种基于dct域的hevc压缩噪声水平估计去噪方法
CN112243126A (zh) 一种结合视频特性的hevc自适应量化参数优化方法
WO2022140905A1 (zh) 预测方法、编码器、解码器以及存储介质
CN113596483B (zh) 一种编码树单元的参数确定方法及系统
CN113395518B (zh) 一种视频编码自适应量化死区的方法
WO2023245544A1 (zh) 编解码方法、码流、编码器、解码器以及存储介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21780648

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21780648

Country of ref document: EP

Kind code of ref document: A1