Joint Collaborative Team on Video Coding (jct-vc)

5.20Non-normative: Encoder optimization, decoder speed improvement, post filtering, loss concealment, rate control

5.20.1Rate control

JCTVC-J0057 Adaptive rate control for HEVC [J. Si, S. Ma, W. Gao (Peking Univ.), M. Yang (Huawei)]

This contribution provides improvements of the rate control scheme proposed in JCTVC-I0433. The proposed rate control scheme implemented on HM7.0 is modified mainly in two aspects. One is that hypothetical reference decoder (HRD) is used to adjust bit allocation. The other is that the quantization scale is used in the R-Q model instead of using QP directly. Compared with the original rate control scheme in HM7.0, the average BD-RATE computed using piece-wise cubic interpolation can be up to -28.5% for RA-main (for LP-main: -21.1%; LB-main: -20.7%).

The scheme shows significantly larger fluctuations in PSNR than the current HM 7 rate control – this indicates quality fluctuations which are likely to be observable (e.g. flicker). This could e.g. be evaluated by comparing standard deviations of PSNR.

• 
  Was the CPB buffer length the same in both RC algorithms?
  
• 
  Was the bit rate variation similar?
  

No action.

JCTVC-J0260 Source code bug fix on the RQ model equation [H. M. Choi, J. H. Nam, D. G. Sim (KWU)]

Presentation not uploaded.

This contribution corrects the RQ model equation in the rate control of HM software.

Tries to correct the too-aggressive change in QP – current scheme has large QP fluctuations.

Decision (SW): Adopt.

JCTVC-J0262 Initial QP set for rate control [H. M. Choi, J. H. Nam, D. G. Sim (KWU)]

This contribution presents a condition of initial QP setting in beginning of a sequence for the rate control implemented on HM7.0 and complexity of the rate control. The initial QP setting condition was already introduced at the last meeting, JCTVC-I0094 includes identical condition. In this document, the identical condition is implemented in HM7.0 and tested. The initial QP is automatically calculated based on bit per pixel (bpp) with target bitrate. Therefore, the QP value in the configuration file is ignored. Furthermore, this proposal provides complexities by rate control in encoding.

Decision (SW): Adopt, but implement as an option, i.e. it should still be possible to disable and set the initial value manually.
JCTVC-J0263 Reports on improved RD performance of rate control [H. M. Choi, J. H. Nam, S. M. Yoo, D. G. Sim (KWU), Y.-H. Kim, J. H. Park (KETI)] [late]

This contribution presents improvement of rate control algorithm based on JCTVC-H0213 and JCTVC-I0094. This document proposes a scheme of boundary clipping for target bits in the frame level rate control. The boundary for target bits of a frame is used to conform to the hypothetical reference decoder (HRD) requirement, so the target bits are clipped by lower and upper bounds. The lower and upper bounds are adaptively derived from the amount of coded picture buffer (CPB). In addition, this document introduces an adaptive MAD prediction on LCU level regarding to correlation of spatial and temporal MAD cost to improve accuracy of MAD prediction.

The contribution was a description of wWork still in progress – no action taken.

5.20.2Tools

JCTVC-J0126 Code for verifying the coding tree block raster, tile and z-scan processes [A. K. Ramasubramonian, Y.-K. Wang (Qualcomm)]

There have been successive corrections made to the pseudocode specifications for the coding tree block raster scan, tile scan and z-scan processes in Subclauses 6.5.1 and 6.5.2. A C-code was created to verify the pseudocodes. After final bug fixes, the code has verified that the processes currently specified in WD7 are now indeed correct. This document provides, for archive purpose, the C-code. The code is provided as an attachment with this document, along with output of a few test cases.

No need to present – informative contribution to make C code available.

5.20.3Encoder optimization

JCTVC-J0242 Encoding with fixed Lagrangian multipliers [J. Xu (Microsoft)]

This contribution presents a fixed Lagrange multiplier encoding method that is applied instead of using fixed QP encoding (as used in HM). When using the proposed fixed Lagrange multiplier encoding algorithm, −1.7%, −1.8%, −1.4%, and −1.4% Y BD-Rate changes are reported for RA-Main, RA-HE10, LD-Main, and LD-HE10, respectively. The bit rate savings for U and V are reportedly somewhat larger. The encoding/decoding times of the proposed fixed Lagrange multiplier encoding algorithm are approximately unchanged. This contribution does not include any normative changes.

This contribution is a follow-up of JCTVC-I0426 for which several questions were raised

The approach is to assign lower QP / higher rate to temporal layer 0, and less quality to other layers.

Typically at one given QP point, the bit rate is higher.

It is reported that the visual quality is not affected.

A version with rate control is also presented, but this is still work in progress.

A version with LCU-based multi-pass is also investigated with additional gain.

Decision (SW): Adopt as encoder option.
JCTVC-J0134 Cross-check of Encoding with Fixed Lagrangian Multipliers (JCTVC-J0242) [Y. Chiu, W. Zhang (??)], J. Zhang (Intel) [late]

5.21To be allocated | category not clear

5.22Withdrawn and missing contributions

JCTVC-J0058 [Withdrawn]
JCTVC-J0117 [Withdrawn]

JCTVC-J0294 [Withdrawn]
JCTVC-J0384 [Withdrawn]
JCTVC-J0396 [Withdrawn]
JCTVC-J0404 [Withdrawn]
JCTVC-J0434 [Withdrawn]
JCTVC-J0441 [Withdrawn]
JCTVC-J0461 [Withdrawn]
JCTVC-J0497 [Withdrawn]