Organisation internationale de normalisation
Yüklə 9,04 Mb.
|
Bitstream exchange
Post-processing notes: For 4:2:0 de-interlacing, use "off-line" de-interlacer so that a progressive YUV sequence can be played back for the subjective test. This avoids any temporal interaction resulting from difficulty in synchronising playback with the built-in hardware de-interlacer in the display. The ffmpeg software has several de-interlacers, and two candidates are suggested to be ‘w3fdif’ and "mcdeint". Recommendations:
It was suggested to produce a draft test plan. 13.0.0.1.1.1.1.1.51JCTVC-S0006 SCC coding performance analysis (AHG6) H. Yu (chair), R. Cohen, A. Duenas, S. Liu, K. Rapaka, J. Xu (vice chairs) This report summarizes the activities of the JCT-VC ad hoc group on SCC coding performance analysis (AHG6) between the JCT-VC 18th meeting in Sapporo, Japan, and the 19th meeting in Strasbourg, France. JCTVC-R1015 “Common conditions for screen content coding tests” was produced based on the agreements captured in the meeting notes. The AHG revised the common test condition described in JCTVC-Q1015 with minimum changes necessary to reflect SCM 2.0 related information. In particular, the following new SCM2.0 encoder settings were used for generating the anchors:
Results reporting templates were updated with the SCM2.0 anchor data. For the alternative IBC operating point with constrained search range defined and specified in the JCTVC-R-Notes_dB document, the related SCM 2.0 macro settings are not included in R1015. However, was a recommendation in email on the JCTVC reflector to use the following configuration for this:
No change was made in JCTVC-R1015 towards analysis methods for studying SCC coding performance. However, comments were made during the last meeting that suggested changes and improvements to the Excel templates for making comparisons among multiple proposals easier, e.g. using bar charts only showing average across text and graphics and mixed content. In fact, bar charts continue being used in a number of proposals this time. It was reported to be desirable to create a unified solution and include such a solution in the test results reporting template. JCTVC-S0042 “AHG6 Inclusion of 420 screen content in the test conditions” proposes that 4:2:0 screen content, such as sequences in Class F sequences should be added to common test condition.
JCTVC S0100 “AHG6: On Adaptive Colour Transform (ACT) in SCM2.0” recommends that ACT should be disabled in common test conditions for YUV sequences. Changes were also made to the encoder to only evaluate ACT on for RGB content, such that the decoded CU-level ACT flags are always 1. It is reported that for coding RGB content, forcing CU-level ACT flags always to be on, preserved most of the gains achieved by CU-adaptive ACT. The main arguments are that from the provided test results, it is asserted that “ACT itself does not provide much benefit for YUV content”, and on RGB content, testing both CU-level on and off provides only 1% gain over only testing CU-level on. The proposal states that both suggestions would significantly reduce simulation runtime, while preserving coding efficiency. Various tests have been designed in the CEs to analyze the coding efficiency as well as the associated complexity of the respective tools under test. In addition, a large number of non-CE contributions to his meeting report coding efficiency gain and provide complexity analysis data. In general, it is recommended that the coding performance of a particular coding tool or mode should be measured and described by both coding efficiency and associated complexity. Several experts have also suggested that the interaction of the newly adopted tools with the existing HEVC tools should be tested. JCTVC-S0042 reports the test results that came from running SCM-2.0 with Class F sequences. It states that coding gain from IBC has been observed. Note however, that the Class F sequences were captured using an analog cable, and therefore the screen-content tools that were designed to work well on perfectly flat or limited-colour blocks may be affected by noise in the analog-captured signal. JCTVC-S0100 reports the results from two experiments designed for testing ACT. In the first test, it is asserted that the effects of ACT on YUV content mostly came from encoder-side bypassing IBC early termination, rather than form ACT itself. In the second test, it is asserted that for RGB content, the CU-level adaptive colour transform only provides about 1% gain over the case where the colour transform is executed for every CU encoded in the bitstreams. JCTVC-S0069 proposes to enable the colour transform for Inter modes at sequence parameters set instead of enabling it at CU level. A BDR average of 0% and −0.1% compared to SCM2.0 for respectively RA and LDB configurations is reported with an encoding run time of 95% and 96%. The AHG recommended to:
In further discussion chaired by GJS on Wednesday 10-22, the following simplified format was suggested for summary tables:
In further discussion 10-23 (JRO), it was agreed that this shall be done. It was remarked that two of the 720p sequences have extremely low bit rates. 13.0.0.1.1.1.1.1.52JCTVC-S0007 JCT-VC AHG report: SCC extensions text editing (AHG7) [R. Joshi, J. Xu (AHG co-chairs), Y. Ye, S. Liu, R. Cohen, Z. Ma (AHG vice-chairs)] This document reports on the work of the JCT-VC ad hoc group on SCC extensions text editing (AHG7) between the 18th JCT-VC meeting in Sapporo, JP (June-July 2014).and the 19th JCT-VC meeting in Strasbourg, FR (October 2014). The first working draft for the High Efficiency Video Coding Screen Content Coding (HEVC SCC) extension was created as an output document following the decisions taken at the 18th JCT-VC meeting in Sapporo, JP (June-July 2014). Three versions of JCTVC-R1005 were published by the editing ad hoc group (AHG) following the 18th JCT-VC meeting in Sapporo, JP, which were based on the seventh draft of HEVC range extension, i.e., JCTVC-Q1005_v9. List of changes with respect to JCTVC-Q1005_v9:
The screen content coding test model 2 (SCM 2) (document JCTVC-R1014) was released on 16th October, 2014. General tutorial descriptions for palette mode, residual adaptive colour transform and intra block copying were added. For the palette mode and the residual adaptive colour transform, the working of the encoder was also described in greater detail. Recommendations:
13.0.0.1.1.1.1.1.53JCTVC-S0008 SCC extensions software development (AHG8) B. Li, K. Rapaka (chairs), R. Cohen, P. Chuang, X. Xiu, M. Xu (vice chairs) This report summarizes the activities of Ad Hoc Group 8 on screen content extensions software (SCM) development that have taken place between the 18th and 19th JCT-VC meetings. Multiple versions of the HM SCM software were produced and SCM2.0 was announced on the JCT-VC email reflector. The integration details and performance summary of these revisions are provided in the next subsections. The performance results of software revisions were observed to be consistent with the adopted techniques. HM15.0_RExt8.0_SCM2.0 has been tagged on the SVN HHI repository on August 1st. This release includes following adoptions:
In addition as per the meeting discussion, the support for flexible intra block copy search range has been provided. This enables CE specific Intra BC search configurations that are different from to CTC. The release was announced on the email reflector. The software can be downloaded at https://hevc.hhi.fraunhofer.de/svn/svn_HEVCSoftware/tags/HM-15.0+RExt-8.0+SCM-2.0/ The performance of this version against HM-14.0+RExt-7.3+SCM-1.1 was described according to the common test conditions in JCTVC-R1015. For the lossy configuration, it is reported that this version provides BD-rate reduction of 33.0%, 23.4% and 18.1% for RGB 1080p text and graphics category in AI/RA/LB configurations respectively and BD-rate reduction of 31.3%, 17.6% and 11.9% for YUV 1080p text and graphics category in AI/RA/LB configuration, respectively. For the lossless configuration, it is reported that this version provides BD-rate reduction of 19.6%, 11.1% and 10.4% for RGB 1080p text and graphics category in AI/RA/LB configurations respectively and BD-rate reduction of 22.6%, 10.8% and 10.0% for YUV 1080p text and graphics category in AI/RA/LB configuration, respectively. Yüklə 9,04 Mb. Dostları ilə paylaş:
|