Back to Search Document details
4th Meeting: Chengdu, October 2016 2016-10-16 07:23
Improvements on AMT for inter prediction residuals
Authors: Xin Zhao Qualcomm Vadim Seregin Qualcomm Marta Karczewicz Qualcomm

Abstract not available in document

Meeting Notes
JVET-D0126 Improvements on AMT for inter prediction residuals [X. Zhao, V. Seregin, M. Karczewicz (Qualcomm)]

Discussed Sunday 1445 (GJS & JRO)

This contribution is related to the Advanced Multiple Transform (AMT) method for inter prediction residuals in JEM-3.0. For 4-pt, 8-pt and 16-pt transforms, it is proposed to replace the current AMT transform cores, i.e., DST-7 and DCT-8, with separable KLTs. As compared to JEM-3.0, with the proposed method, the overall BD-rate gain for luma component is reportedly 0.12%, 0.35% and 0.34% for Random Access (RA), Low Delay B (LDB) and Low Delay P (LDP) configurations, respectively.

This propoposal provides relative low gain, not justifying to add more transform kernels.

No interest was expressed by other experts.

Cited:
JVET-D0001 JVET-J0021 JVET-J0029 JVET-J1026 JVET-K0356 X JVET-K0375 JVET-Z0147 JVET-Z0210
ARCHIVE v3 :
JVET-D0126.doc JVET-D0126.pdf JVET-D0126r1.xlsm
Decisions
No interest was expressed by other experts.
Citation