This method is mainly used to improve the bit rate control performance of video coding. The proposed method is submitted in H.265 and H.266 respectively, involving proposals JCTVC-AH0024 and JVET-K0390.

Skipped CTU accounted for the majority in the high resolution video, resulting in small average BPP (bits per pixel) and flat RD curve. However, the RD curve of non-skip CTU varies greatly, so it differs greatly from the overall RD curve. In the default bit-rate control framework, the CTU level lambda and QP are truncated by the frame-level lambda and QP, which causes the non-skip CTU encoding parameters to be incorrectly truncated. In order to solve this problem, the proposal proposes to consider skip CTU and non-Skip CTU respectively, and only use data in non-Skip CTU to update frame-level coding parameters.

Parameter update strategy:

1. Obtain the average bits and lambda of a non-Skip CTU

2. Calculate the BPP of a non-Skip CTU area using average bits

3. Use the calculated BPP and lambda to update RD model parameters

Bit rate control with extended RA configuration:

In RA configuration, the default bit rate control only supports GOP=8, while in RA mode the GOP can reach 16. The main problem lies in the target bit allocation. The current image-level bit rate allocation allocates 8 frames of images according to the weights in the following table.

 

In this method, the bit ratio and QP offset of each layer can be obtained according to the method in JCT-VC W0062 and JCT-VC X0038. According to the QP offset and the relationship between QP and lambda, the ratio of lambda of different layers can be obtained. Then, the target bit of each frame can be calculated by the following formula:

 

This method is only suitable for the situation where the initial QP is known. When the first two GOP are encoded, the initial QP value cannot be obtained and can only be calculated by the original method.

Experimental results:

Experimental determination in HM16.20 LDB and RA mode, YUV BD – Rate is respectively 2.3% / 0.8% / 0.5% and – 7.8% / 6.7% / 6.4%

Table 1 and 2 are the experimental results compared with the default bit rate control in LDB and RA mode, and Table 3 and 4 are the experimental results compared with the fixed QP.

Reference:

JCTVC-AH0024-v1 Improvements for HEVC rate control

JVET-K0390-v5 Rate Control for VVC

Related proposals to download:

Link: pan.baidu.com/s/1aVLwk42L…

Extraction code: XZOA

Copy this section of content after opening Baidu web disk mobile App, more convenient operation oh

Please pay attention to wechat public account Video Coding communication