Results of the current CE do not currently indicate a need for inclusion of any of the additional elements concerning tools 1-5.
Continue CE7 on:
-
Secondary transforms (JCTVC-F224, JCTVC-F554, JCTVC-F294) – implication of using a secondary transform must be more thoroughly assessed, to be able to judge whether an expected 0.7% BR reduction is worthwhile.
-
Extension of DST/DCT for chroma (JCTVC-F553)
5.8.1Summary
5.8.1.1.1.1.1.1.1JCTVC-F028 CE8: Summary report of Non-deblocking Loop Filtering [T. Yamakage, I. S. Chong, M. Narroschke] [upload 07-12]
A summary of core experiment 8 (CE8) on Non-deblocking loop filtering was reported. There are five Subtests in CE8,
-
Subtest1: Enhance block level filter adaptation with directional features (three proposals, software checked),
-
Subtest 2: Effectiveness of adaptive loop filtering scheme are evaluated for using fixed set of filters (one proposal),
-
Subtest 3: Memory reduction schemes in ALF decoding (one proposal),
-
Subtest 4: Complexity reduction and/or quality improvement (three proposals)
-
Subtest 5: Effectiveness of chroma adaptive loop filtering scheme (two proposals).
These were evaluated based on the common conditions in JCTVC-E700 and additional conditions in JCTVC-E708. All mandatory results are verified by cross-checkers.
"Subtest 0" / "Homework" ALF on vs. ALF off comparison, as tabulated below:
|
|
High Efficiency
|
|
AI
|
RA
|
LDB
|
LDP
|
BD-rate
|
Class A
|
4.1
|
6.7
|
|
|
Class B
|
2.3
|
4.9
|
5.5
|
8.7
|
Class C
|
2.1
|
3.0
|
3.7
|
3.8
|
Class D
|
0.9
|
2.8
|
2.5
|
1.7
|
Class E
|
3.5
|
|
7.7
|
11.5
|
Overall
|
2.5
|
4.4
|
4.7
|
6.2
|
Enc. Time
|
82
|
94
|
95
|
93
|
Dec. time
|
75
|
81
|
81
|
79
|
Informally (very informally), it was reported that ALF seemed to never hurt subjective quality, and to sometimes substantially improve subjective quality.
Dostları ilə paylaş: |
