Organisation internationale de normalisation
| CS1
| ||||||
|
CS2 | ||||||
|
|
Y |
U |
V |
Y |
U |
V |
|
1920x1080p |
-16.5% |
-23.0% |
-23.2% |
-13.9% |
-20.5% |
-20.0% |
|
832x480p |
-7.4% |
-5.7% |
-3.4% |
-4.0% |
3.8% |
8.3% |
|
1280x720p |
-12.6% |
-14.3% |
-16.8% |
-0.5% |
-4.8% |
-7.1% |
|
Overall |
-12.1% |
-14.3% |
-14.3% |
-6.6% |
-7.4% |
-6.2% |
For CS1, IVC outperforms WVC in almost all sequences except the sequence “racehorces”
For CS2, IVC outperforms WVC in most of all sequences except the three sequences “basketballdrill”, “fourpeople” and “KristenAndSara”.
13.0.0.1.1.1.1.1.32m35351 Cross-check for performance evaluation of ITM 10.0 over WebVC according to the VCC condition (m35003) [Anna Yang, Soo-Chang Oh, Dong-Hyun Kim, Jae-Gon Kim]
13.0.0.1.1.1.1.1.33m35004 Comments on WD3 of Internet Video Coding (IVC) [Jin Yeon Choi, Sang-hyo Park, Euee S. Jang]
Abstract: Presents editorial and technical comments on WD3 of IVC to improve the quality and leave no room for confusion.
BoG recommendation: Adopt the following 10 Editorial comments for next version of WD:
(1) The elements of Subclause 3.1 (Terms and definitions) are sorted in alphabetical order.
(2) Add the elements of “abt_enable” and “if_type” to sequenceheader
(3) There are no the descriptions for “abt_enable” and “if_type” (i.e., no semantics for the elements).
(4) Remove “S pictures”
(5) Fix Figure C-3.
(6) Refine “Backward prediction” definition
(7) Refine the term of “encoding process”
(8) Divide “Definitions” into multiple sections
(9) Divide “Bitstream syntax and semantics” into two sections
(10) Remove descriptors that are not used in the WD3
13.0.0.1.1.1.1.1.34m35006 Report on the entropy of prediction types in IVC [Myeong Kyun Kim, Sang-hyo Park, Euee S. Jang]
(Withdrawn.)
13.0.0.1.1.1.1.1.35m35053 Performance evaluation of internet video coding [Xufeng Li, Ronggang Wang, Zhenyu Wang, Wenmin Wang, Siwei Ma, Tiejun Huang, Wen Gao]
In the RA constraint cases (CS1), IVC (as represented by the ITM software encoder) outperforms WVC (as represented by the JM reference software encoder) in terms of BD bit rate numbers in overall average. It is quite apparent for the 1280x720p sequences. But for Sequence RaceHoreses (416x240), IVC ITM encoder underperforms WVC.
In the LD constraint cases (CS2), the WVC encoder clearly falls behind IVC ITM encoder in terms of BD bit rate. The IVC ITM encoder has 7.7% lower in bit rate on average than WVC respectively. For Sequence Kimono, PartyScene and BQSquare, the IVC ITM encoder has more than 15% lower in bit rate than WVC encoder.
The B frame number should be fixed, and it is expected to be discussed the common test condition of IVC during the following days to produce common test results.For example, the B frame numbers, the QP values, and other parameters.
ITM10 outperforms ITM9 by 8-9% both for CS1 and CS2 cases.
13.0.0.1.1.1.1.1.36m35054 De-blocking improvement for internet video coding [Xufeng Li, Ronggang Wang, Zhenyu Wang, Wenmin Wang, Siwei Ma, Tiejun Huang, Wen Gao]
This contribution proposes to add a loop filtering tool in the coding loop of Internet Video Coding Test Model. Experimental results show that the proposed scheme achieves about 1.1% and 0.8% bit rate saving in CS2, CS1 configure respectively and subjective quality has improved significantly.
There is some subjective quality improvement by the new filter.
Proposes to add a strong filter beside the existed nomal filter. The strong filter is designed to correct the pixel value according to the portion of the difference of inter-block pixels, and the proposed method is said to be covered by the claims of the following expired patent: United States Patent 5337088, Honjo, M., “Method of correcting an image signal decoded in block units”, 1993.
BoG recommendation: Adopt it in the next version of ITM, when other new tools for the new version of ITM are ready, the proposed method should be retested to verify the performance gain.
13.0.0.1.1.1.1.1.37m35353 Cross-check of m35054 (De-blocking improvement for internet video coding) [Jin Yeon Choi, Sang-hyo Park, Euee S. Jang]
13.0.0.1.1.1.1.1.38m35056 4x4 blocksize transform for internet video coding [Zhenyu Wang, Xufeng Li, Ronggang Wang, Siwei Ma, Tiejun Huang, Wen Gao]
Transform with 4x4 block size is an effective tool for video coding. In edge regions, intra prediction and transform with 4x4 block size can provide significant performance gain and improve subjective quality on IVC platform. This contribution proposed to add 4x4 transform to intra block of IVC. Experiments on ITM10 show that the average BD rate saving is 0.8% in CS2, and 2.6% in CS1.
Propose to add 4x4 transform and intra prediction to the intra block.
The 4x4 transform can also help to alleviate ringing artifacts in low bitrate scenario.
The ideas of “quadtree-based ABT” and integer transform are published 20 years ago.
The direction spatial intra prediction technique of 4x4 block is first disclosed in Q15-F-11, “Response to Call for Proposals for H.26L" by Gisle Bjontegaard on Oct. 1998. No related patent has been filed by Gisle Bjontegaard on intra prediction.
BoG recommendation: to adopt it in the next version of ITM.
13.0.0.1.1.1.1.1.39m35350 Cross-check for 4x4 Transform for IVC (m35056) [Anna Yang, Soo-Chang Oh, Dong-Hyun Kim, Jae-Gon Kim]
- Yüklə 9,04 Mb.
Dostları ilə paylaş:
