Study of this topic remained ongoing, with non-camera content coding being considered in range extensions development and also in preparations for a future call for proposals focusing specifically on the coding of this type of content as its primary requirement.
HEVC for Interlaced Scan Video
The following related input contributions were noted. See the Requirements report for plans in this area.
14.1.1.1.1.1.1.1.82m31286 Proposal of interlace coding tools for HEVC [G.Barroux, K.Kazui, A.Nakagawa(Fujitsu)]
14.1.1.1.1.1.1.1.83m31616 On the cross-check of m31286 on proposed interlace coding tools [D. Baylon, K. Panusopone, A. Luthra (ARRIS)]
14.1.1.1.1.1.1.1.84m31659 Cross check of low-delay results for study of interlace [Greg Coppa, Greg Rutz, Arianne Hinds]
Use cases
High-dynamic range and extended color gamut content coding remained a subject of study for potential additional future work. The contribution below was noted to be related to this area. See the Requirements report for plans in this area.
14.1.1.1.1.1.1.1.85m31308 The adaptive LogLUV transfrom for mapping HDR video to traditional video codecs [Herbert Thoma]
-
Further study of additional needs for HEVC application to still picture coding remains of interest. See the Requirements report for plans in this area.
Wednesday Video plenary status review
The plenary was held 11:00-13:00 on Wednesday. Reports of BoGs were given and further discussed as follows.
CDVS:
-
Initial decision on new method of keypoint detection (denoted as ALP)
-
Preparation of CD
-
Preparation of CEs
IVC:
-
Further review necessary on prior art for new proposals (e.g. intra smoothing)
-
Viewing sessions to be organized – get volunteers
WVC:
-
Comment of US on normatively referencing 14496-10 – it is best to resolve this by copying the respective parts (e.g. SEI messages) into 14496-29
-
More work necessary on this
-
Not all patent statements have arrived yet, more study of situation necessary
-
Suggestion to delay for one meeting cycle
-
Issue Draft DoC and Draft FDIS.
-
Responses to non-US NB cannot be drafted yet.
VCB:
-
Set up reference test conditions, and generate bit streams of the current VCB as anchors that would allow other experts to bring in and show the merit of new technology
-
Bit streams of CfP may not fulfill this purpose, there were many cases where target rates were not met
-
Establish conditions that allow assessing the merit of additions. One suggestion was that fixed QP be used (within frame classes), since rate control might interfere with merit of additions, quality should be constant over sequence
General requirements for IVC, WVC and VCB were discussed.
It was agreed to start an effort to perform a visual comparison between the different codecs to gain more information about their performance, and to assess their other technical merits.
JNB comment needs to be taken into account, identifying the justification of possibly standardizing various technologies – this would be resolved by such an action, gathering more information
Complexity and implementation analysis should also be included.
Further discussion on this aspect Thu 10:00:
Some comments that were made in the discussion
-
Desire was expressed to also investigate other issues, e.g. behaviour under channel errors, bandwidth fluctuations
-
This extended analysis would be desirable, but may be too complicated as a first step, investigation of “plain”
-
Suggestion was made to produce for each of the codecs bitstreams with sufficiently dense rate points within the range of CFP rates, such that visual comparison is enabled with sufficient close rate points. By the next meeting, rate points (e.g. low, medium high) which are sufficiently clodse would be selected. If one of the codecs is not able to meet that rate point, it is useless for comparison. Agreed.
-
Some experts requested that the results of such an effort might be made public. There is no agreement about this at this point in time, and no need to discuss it now.
-
Several experts expressed opinion that influence of rate control should be excluded. Arguments brought for this were a) bit rate should not be shifted between parts of the sequence b) for comparing short clips, quality should not vary
-
Fixed QP overall would not be a solution, as it unreasonably gives up quality of certain tools
-
It was suggested that a “shootout” could be used, i.e. handle encoders as black box – there was no agreement to do this, as it would make the interpretation difficult.
It was decided to issue output document “Conditions for visual comparison of VCB, IVC and WVC codecs” which should contain the following sentences as a basis
-
produce for each of the codecs bit streams with sufficiently dense rate points within the range of recent CFP rates (tbd during editing period), such that visual comparison is enabled with sufficient close rate points. By the next meeting, rate points (e.g. low, medium high) which are sufficiently close would be selected. If one of the codecs is not able to meet that rate point, it is useless for comparison.
Agreed constraints for the intended visual comparison of codecs:
-
Allow QP variation within a sequence within a pattern of frame types (e.g. I/B/P, normal/golden), but limit quantizer step size ratio to a range of (tbd during editing period), same ratio settings shall be used for all sequences by a specific codec.
-
No adaptation of the GOP pattern per sequence
-
Frequency of intra pictures should be identical as far as possible in RA scenario, and not supersede the value of recent CfP
-
Disallow rate control.
A first version of this document was edited by the video chair, and approved during the plenary on Thursday afternoon. The document requires further improvements which are to be finalized within 2 weeks after the meeting. An AHG N14110) was set up, with the mandate to further work on the preparation of the viewing tests to be performed during the next meeting.
The VCB AHG shall also take on the task of investigating a set of common test conditions and a set of anchors that can be used for proposals on enhancements of VCB. As a first example, the effect of rate control (testing once with the previous CfP settings, and once with the “constrained quantizer” settings as from the visual test).
Dostları ilə paylaş: |