Organisation internationale de normalisation


Residual downsampling for MVC



Yüklə 2,74 Mb.
səhifə71/96
tarix02.01.2022
ölçüsü2,74 Mb.
#28941
1   ...   67   68   69   70   71   72   73   74   ...   96

5.5Residual downsampling for MVC




5.5.1.1.1JVT-Z023 ( Prop 2.2) [S. Cho, N. Hur, J. Kim, S.-I. Lee (ETRI)] Coding eff of stereoscopic video coding using residual downsampling

This contribution reported an analysis of experimental results of stereoscopic video coding using the residual-downsampling algorithm which was contributed in JVT-Y052 at the last meeting. It also compared the coding efficiency of the residual-downsampling algorithm in JVT-Y052 with that of the JMVM including 8x8 transforms for only the non-base view.
The basic concept is analogous to RRU of H.263 Annex Q / MPEG-4 pt. 2, plus also supporting horizontal-only and vertical-only downsampling.
Tested with CAVLC, IPPP coding structure, no 8x8 transform in non-base view.
Tested on three non-common-conditions test sequences, one of which was animation content.
Significant coding gain was reported at lower bit rates (with or without 8x8 transform in non-base view).
With the 8x8 transform in both the reference and modified design, the contribution reports significant gain for the use of RRU: About 8% on two natural video sequences, and 16% on one animation sequence, as a percentage of the bit rate for the second view.
Remark: But this is the percentage savings of just the bits for one of the two views – not of the total bit rate. Response: Yes, as a percentage of the total, it would be about 3% for the natural sequences and perhaps 5% for the animation.
Remark: Does this make sense? How would it perform on regular single-view video?
Remark: Common conditions sequences are not stereoscopic sequence.
Remark: Is there really anything about this proposal that makes it specific to the stereoscopic context? Perhaps not.
Remark: JMVM with and without 8x8 transform – no significant difference in performance was reported between these two. Response: Probably because this is using QVGA resolution sequences.
Not verified, not using common conditions, just a few (low-res) test sequences.
Remark: Should test using MVC common conditions.
JVT Disposition: Further study was encouraged.


Yüklə 2,74 Mb.

Dostları ilə paylaş:
1   ...   67   68   69   70   71   72   73   74   ...   96




Verilənlər bazası müəlliflik hüququ ilə müdafiə olunur ©muhaz.org 2024
rəhbərliyinə müraciət

gir | qeydiyyatdan keç
    Ana səhifə


yükləyin