Back to Search Document details
33rd Meeting: by teleconference, CH, January 2024 2024-02-27 03:09
Exploration experiment on enhanced compression beyond VVC capability (EE2)
Abstract
This document describes Exploration Experiments (EEs) planned to be performed between 33rd and 34th JVET meetings to evaluate enhanced compression tools beyond VVC capability.
JVET-AG2024 Exploration experiment on enhanced compression beyond VVC capability (EE2) [V. Seregin, J. Chen, R. Chernyak, K. Naser, J. Ström, F. Wang, M. Winken, X. Xiu, K. Zhang] [WG 5 N 267] (2024-02-23)

An initial draft of this document was reviewed and approved at 0730-0735 on Friday 26 Jan.

Categories and experiments are listed in the following table:

Tests

Tester

Cross-checker

1 Partitioning

1.1a

Temporal prediction of split modes

G. Laroche

(Canon)

1.1b

Test 1.1a + Maximum MTT depth adaptation

G. Laroche

(Canon)

1.1c

Test 1.1b + Prediction of split syntax elements order

G. Laroche

(Canon)

2 Intra prediction

2.1

TIMD merge mode

R. Youvalari (Xiaomi)

Y. Wang

(Bytedance)

2.2

Occurrence-based intra coding

R. Youvalari (Xiaomi)

2.3a

Intra prediction using merged histogram of gradients

S. Blasi

(Nokia)

2.3b

Intra prediction using merged histogram of gradients (without restrictions)

S. Blasi

(Nokia)

2.4

DIMD merge list

M. Blestel

(Ofinno)

2.5a

Test 2.3a + Test 2.4

S. Blasi

(Nokia)

M. Blestel

(Ofinno)

2.5b

Test 2.3a + Test 2.4 + Test 2.2

R. Youvalari (Xiaomi)

S. Blasi

(Nokia)

M. Blestel

(Ofinno)

2.6a

Adaptive MRL fusion with fixed weights

S. Blasi

(Nokia)

2.6b

Adaptive MRL fusion with adaptive weights

S. Blasi

(Nokia)

2.7

Relaxing line buffer restriction

Z. Deng

(Bytedance)

2.8

Test 2.6b + Test 2.7

Z. Deng

(Bytedance)

S. Blasi

(Nokia)

2.9a

EIP with bias

J. Lainema
(Nokia)

2.9b

EIP with clipping

J. Lainema
(Nokia)

2.9c

EIP with bias and clipping

J. Lainema
(Nokia)

2.10a

SATD based IntraTMP BV search without signalling

W. Chen

(Kwai)

2.10b

SATD based IntraTMP BV search with signalling

K. Naser

(InterDigital)

2.10c

MR-SAD based IntraTMP BV search with signalling

K. Naser

(InterDigital)

2.10d

IntraTMP BV reordering

W. Chen

(Kwai)

2.10e

Test 2.10a + Test 2.10d

W. Chen

(Kwai)

2.10f

Test 2.10b + Test 2.10d

K. Naser

(InterDigital)

W. Chen

(Kwai)

2.11a

Auto-relocated BVP for IntraTMP merge candidates

L. Zhang

(OPPO)

N. Qiu

(Xidian Univ.)

2.11b

Template matching based sorting of IntraTMP merge candidates with maximum list size of N

L. Zhang

(OPPO)

N. Qiu

(Xidian Univ.)

2.11c

Test 2.11a + Test 2.11b

L. Zhang

(OPPO)

N. Qiu

(Xidian Univ.)

2.11d

Test 2.11c + Restricting the BV range for IntraTMP merge candidates

L. Zhang

(OPPO)

N. Qiu

(Xidian Univ.)

2.12a

Chroma mode reordering based on unextended list

X. Li

(Alibaba)

2.12b

Chroma mode reordering based on extended list

X. Li

(Alibaba)

2.13

Matrix based intra prediction replacing conventional intra modes

B. Ray

(Qualcomm)

2.14a

Bilateral filtering for intra prediction (not applied to planar mode)

W. Yin

(Bytedance)

2.14b

Bilateral filtering for intra prediction (applied to planar mode)

W. Yin

(Bytedance)

3 Inter prediction

3.1a

Chained motion vector prediction

Y. Kidani (KDDI)

3.1b

Test 3.1a + Restriction of only motion from collocated pictures

Y. Kidani (KDDI)

3.2a

Removing the one-CTU-row temporal buffer constraint for all relevant tools

Z. Deng

(Bytedance)

3.2b

Imposing the one-CTU-row temporal buffer constraint on all relevant tools

Z. Deng

(Bytedance)

3.2c

Test 3.2a + Test 3.1a

Z. Deng

(Bytedance)

Y. Kidani (KDDI)

3.2d

Test 3.2b + Test 3.1a

Z. Deng

(Bytedance)

Y. Kidani (KDDI)

3.3

Inter CCP merge mode with zero luma CBF

Z. Deng

(Bytedance)

3.4

Adaptive GPM blending

L. Zhao

(Bytedance)

3.5a

LIC model inheritance for GPM

C.-C. Chen

(Qualcomm)

3.5b

Test 3.5a + LIC model inheritance for merge modes at LIC flag swapping stage

C.-C. Chen

(Qualcomm)

3.5c

Test 3.5a + LIC model inheritance for merge modes at ARMC stage

C.-C. Chen

(Qualcomm)

3.6a

Reference filtering for inter-prediction (with reference sample classification into 2 groups)

V. Rufitskiy (Ofinno)

 

3.6b

Reference filtering for inter-prediction (with reference sample classification into the extended number of groups)

V. Rufitskiy (Ofinno)

 

3.7a

TM-based motion refinement with more candidates for SbTMVP

J. Chen

(Alibaba)

3.7b

TM-based motion refinement for affine candidates

J. Chen

(Alibaba)

3.7c

Test 3.7a + Test 3.7b

J. Chen (Alibaba)

4 Transform and coefficients coding

4.1

Multiple transform set selection for LFNST/NSPT

F. Wang

(OPPO)

4.2

Fix on LFNST/NSPT index signalling for inter coding

M. Koo

(LGE)

4.3

16 states TCQ

M. Balcilar

(InterDigital)

5 In-loop filtering

5.1a

Adaptive precision for CCALF coefficients

N. Hu

(Qualcomm)

N. Song

(OPPO)

5.1b

Adaptive precision for CCALF coefficients with precision used for luma ALF coefficients

N. Hu

(Qualcomm)

N. Song

(OPPO)

5.1c

Removal of power of 2 constrain for CCALF coefficients

N. Hu

(Qualcomm)

5.1d

Test 5.1a + Test 5.1c

N. Hu

(Qualcomm)

N. Song

(OPPO)

5.1e

Test 5.1b + Test 5.1c

N. Hu

(Qualcomm)

N. Song

(OPPO)

5.2a

Coding information based classification for ALF (with partitioning info only)

W. Yin

(Bytedance)

N. Hu

(Qualcomm)

5.2b

Coding information based classification for ALF (with prediction info only)

W. Yin

(Bytedance)

N. Hu

(Qualcomm)

5.2c

Test 5.2a + Test 5.2b

W. Yin

(Bytedance)

N. Hu

(Qualcomm)

6 Entropy coding

6.1a

Switching of context initialization for B-slice at slice level

R.-L. Liao

(Alibaba)

6.1b

Switching of context initialization for B-slice at sequence level

R.-L. Liao

(Alibaba)

6.2a

Extended entropy coding with non equiprobable coding

F. Galpin (InterDigital)

6.2b

Test 6.2a but trained on another dataset

F. Galpin (InterDigital)

6.3

Test 6.1 + Test 6.2

F. Galpin (InterDigital)

R.-L. Liao

(Alibaba)

JVET-AG2024 Exploration experiment on enhanced compression beyond VVC capability (EE2) [V. Seregin, J. Chen, R. Chernyak, K. Naser, J. Ström, F. Wang, M. Winken, X. Xiu, K. Zhang] [WG 5 N 267] (2024-02-23)

An initial draft of this document was reviewed and approved at 0730-0735 on Friday 26 Jan.

Categories and experiments are listed in the following table:

Tests

Tester

Cross-checker

1 Partitioning

1.1a

Temporal prediction of split modes

G. Laroche

(Canon)

1.1b

Test 1.1a + Maximum MTT depth adaptation

G. Laroche

(Canon)

1.1c

Test 1.1b + Prediction of split syntax elements order

G. Laroche

(Canon)

2 Intra prediction

2.1

TIMD merge mode

R. Youvalari (Xiaomi)

Y. Wang

(Bytedance)

2.2

Occurrence-based intra coding

R. Youvalari (Xiaomi)

2.3a

Intra prediction using merged histogram of gradients

S. Blasi

(Nokia)

2.3b

Intra prediction using merged histogram of gradients (without restrictions)

S. Blasi

(Nokia)

2.4

DIMD merge list

M. Blestel

(Ofinno)

2.5a

Test 2.3a + Test 2.4

S. Blasi

(Nokia)

M. Blestel

(Ofinno)

2.5b

Test 2.3a + Test 2.4 + Test 2.2

R. Youvalari (Xiaomi)

S. Blasi

(Nokia)

M. Blestel

(Ofinno)

2.6a

Adaptive MRL fusion with fixed weights

S. Blasi

(Nokia)

2.6b

Adaptive MRL fusion with adaptive weights

S. Blasi

(Nokia)

2.7

Relaxing line buffer restriction

Z. Deng

(Bytedance)

2.8

Test 2.6b + Test 2.7

Z. Deng

(Bytedance)

S. Blasi

(Nokia)

2.9a

EIP with bias

J. Lainema(Nokia)

2.9b

EIP with clipping

J. Lainema(Nokia)

2.9c

EIP with bias and clipping

J. Lainema(Nokia)

2.10a

SATD based IntraTMP BV search without signalling

W. Chen

(Kwai)

2.10b

SATD based IntraTMP BV search with signalling

K. Naser

(InterDigital)

2.10c

MR-SAD based IntraTMP BV search with signalling

K. Naser

(InterDigital)

2.10d

IntraTMP BV reordering

W. Chen

(Kwai)

2.10e

Test 2.10a + Test 2.10d

W. Chen

(Kwai)

2.10f

Test 2.10b + Test 2.10d

K. Naser

(InterDigital)

W. Chen

(Kwai)

2.11a

Auto-relocated BVP for IntraTMP merge candidates

L. Zhang

(OPPO)

N. Qiu

(Xidian Univ.)

2.11b

Template matching based sorting of IntraTMP merge candidates with maximum list size of N

L. Zhang

(OPPO)

N. Qiu

(Xidian Univ.)

2.11c

Test 2.11a + Test 2.11b

L. Zhang

(OPPO)

N. Qiu

(Xidian Univ.)

2.11d

Test 2.11c + Restricting the BV range for IntraTMP merge candidates

L. Zhang

(OPPO)

N. Qiu

(Xidian Univ.)

2.12a

Chroma mode reordering based on unextended list

X. Li

(Alibaba)

2.12b

Chroma mode reordering based on extended list

X. Li

(Alibaba)

2.13

Matrix based intra prediction replacing conventional intra modes

B. Ray

(Qualcomm)

2.14a

Bilateral filtering for intra prediction (not applied to planar mode)

W. Yin

(Bytedance)

2.14b

Bilateral filtering for intra prediction (applied to planar mode)

W. Yin

(Bytedance)

3 Inter prediction

3.1a

Chained motion vector prediction

Y. Kidani (KDDI)

3.1b

Test 3.1a + Restriction of only motion from collocated pictures

Y. Kidani (KDDI)

3.2a

Removing the one-CTU-row temporal buffer constraint for all relevant tools

Z. Deng

(Bytedance)

3.2b

Imposing the one-CTU-row temporal buffer constraint on all relevant tools

Z. Deng

(Bytedance)

3.2c

Test 3.2a + Test 3.1a

Z. Deng

(Bytedance)

Y. Kidani (KDDI)

3.2d

Test 3.2b + Test 3.1a

Z. Deng

(Bytedance)

Y. Kidani (KDDI)

3.3

Inter CCP merge mode with zero luma CBF

Z. Deng

(Bytedance)

3.4

Adaptive GPM blending

L. Zhao

(Bytedance)

3.5a

LIC model inheritance for GPM

C.-C. Chen

(Qualcomm)

3.5b

Test 3.5a + LIC model inheritance for merge modes at LIC flag swapping stage

C.-C. Chen

(Qualcomm)

3.5c

Test 3.5a + LIC model inheritance for merge modes at ARMC stage

C.-C. Chen

(Qualcomm)

3.6a

Reference filtering for inter-prediction (with reference sample classification into 2 groups)

V. Rufitskiy (Ofinno)

3.6b

Reference filtering for inter-prediction (with reference sample classification into the extended number of groups)

V. Rufitskiy (Ofinno)

3.7a

TM-based motion refinement with more candidates for SbTMVP

J. Chen

(Alibaba)

3.7b

TM-based motion refinement for affine candidates

J. Chen

(Alibaba)

3.7c

Test 3.7a + Test 3.7b

J. Chen (Alibaba)

4 Transform and coefficients coding

4.1

Multiple transform set selection for LFNST/NSPT

F. Wang

(OPPO)

4.2

Fix on LFNST/NSPT index signalling for inter coding

M. Koo

(LGE)

4.3

16 states TCQ

M. Balcilar

(InterDigital)

5 In-loop filtering

5.1a

Adaptive precision for CCALF coefficients

N. Hu

(Qualcomm)

N. Song

(OPPO)

5.1b

Adaptive precision for CCALF coefficients with precision used for luma ALF coefficients

N. Hu

(Qualcomm)

N. Song

(OPPO)

5.1c

Removal of power of 2 constrain for CCALF coefficients

N. Hu

(Qualcomm)

5.1d

Test 5.1a + Test 5.1c

N. Hu

(Qualcomm)

N. Song

(OPPO)

5.1e

Test 5.1b + Test 5.1c

N. Hu

(Qualcomm)

N. Song

(OPPO)

5.2a

Coding information based classification for ALF (with partitioning info only)

W. Yin

(Bytedance)

N. Hu

(Qualcomm)

5.2b

Coding information based classification for ALF (with prediction info only)

W. Yin

(Bytedance)

N. Hu

(Qualcomm)

5.2c

Test 5.2a + Test 5.2b

W. Yin

(Bytedance)

N. Hu

(Qualcomm)

6 Entropy coding

6.1a

Switching of context initialization for B-slice at slice level

R.-L. Liao

(Alibaba)

6.1b

Switching of context initialization for B-slice at sequence level

R.-L. Liao

(Alibaba)

6.2a

Extended entropy coding with non equiprobable coding

F. Galpin (InterDigital)

6.2b

Test 6.2a but trained on another dataset

F. Galpin (InterDigital)

6.3

Test 6.1 + Test 6.2

F. Galpin (InterDigital)

R.-L. Liao

(Alibaba)

Decisions
JVET-AG2024 Exploration experiment on enhanced compression beyond VVC capability (EE2), also issued as WG 5 N 267
(Alibaba)
Citation