Explorations Media Processing with Privacy
In the actual use of MPEG technologies, there are many contexts that require the processing of the media to be private. Examples of this are searching an encrypted audio visual database with an encrypted query (Figure 1); identify a spoken keyword in a private conversation, e.g. encrypted audio; removal of identification clues from media such as video, audio or speech; sharing multimedia content in a limited context, e.g. make a picture available to a limited list of persons or for a limited time. For details, please see N14285 Processing and Sharing of Media under User Control. The Requirements subgroup and the adhoc group N14286 Adhoc Group on Processing and Sharing of Media under User Control will investigate whether MPEG can provide support for these application scenarios.
Figure 1 Media Processing with Privacy
MPEG User Description
A standardized user description may enable customers to have easier access to services. An ontology and schemas need to be standardized. Information for user description includes attributes, activities, interests, status, history, medical data, etc. User Descriptions are used in MPEG-7, MPEG-21, MPEG-M, and MPEG-V.
MPEG decided at the 105th meeting to issue N13879 Call for Proposals on MPEG User Description (MPEG-UD). At the 106th meeting MPEG received five responses only. These responses provide an uneven coverage of the different requirements. No input from a major service provider was received. The exploration phase ended at the 106th meeting. However, the evaluation report N14287 Evaluation report of the responses to the CfP on MPEG User Description only became available at this meeting. The Systems subgroup may develop a useful standard in the future.
Additional Support for Coding of Interlaced Video in HEVC
At this point, HEVC is capable of coding interlaced material. SEI-messages allow for signalling of interlaced material. Compared to AVC using sequence adaptive frame/field (SAFF) coding, HEVC saves 30% in bitrate for the same objective video quality. If the same subjective video quality is required, HEVC saves 50% of the bitrate compared to AVC. At this point, the HEVC reference software does not properly implement SAFF. This issue has been raised for several meetings now. Furthermore, rate control for coding interlaced video appears to be broken. Unfortunately, nobody undertook a serious effort to mitigate these short comings of the reference software.
At this point, experiments by one company show that relatively small changes to HEVC (colour motion vector scaling, picture-adaptive frame/field coding named PAFF …) can reduce the bitrate by 8% to 15% as shown using selected interlaced material. At this meeting, video sequence were shown on consumer type displays. Unfortunately, a comparison of HEVC with and without the proposed tools at the same bitrate was not shown. Hence, at this meeting no sufficient evidence for issuing a CfP was provided.
The group agreed to use eight test sequence. Six were available at this meeting. One of these sequences is contentious because it was derived from a progressive sequence and includes 3:2 pull-down artefacts. The chair of the test subgroup is to provide two additional interlaced sequences. The group agreed to make available new anchors for the evaluation by February 21st, 2014.
At the 108th meeting, the group with the assistance of the chair of the Test subgroup will make another effort in order to arrive at conclusive results. Given the state of the reference software, this might also involve improving said software. MPEGs position was communicated to VCEG in N14307 Liaison statement to ITU-T SG 16 on video coding collaboration.
Work continues in the N14166 AHG on Study of interlace coding in HEVC. N14176 Requirements and use cases on interlace video and N14175 Draft Call for Proposals for Interlaced Video Coding progress towards a CfP, however subjective evidence has to be shown prior to issuing a CfP.
Coding of medical visual content
Medical images require typically subjectively or mathematically lossless coding. At this point, the Requirements subgroup recommends to wait for the performance evaluation of the current range extensions. In case these extensions do not fulfil the requirements defined in N14173 Requirements for an extension of HEVC for the coding of medical visual content, this exploration will restart with the currently available test material. VCEG was informed of MPEG’s position in N14307 Liaison statement to ITU-T SG 16 on video coding collaboration.
Coding of screen content
Several contributions on screen content coding show that bit rate savings of 35% compared to HEVC Rext 4.1 on pure screen content are possible. The group prepared N14174 Requirements for an extension of HEVC for coding of screen content as well as N14175 Call for Proposals for coding of screen content. This call is joint with VCEG. MPEG decided to have a formal subjective test given the focus on a new type of content and potentially new coding artifacts.
At this and the previous meeting, several profiles for screen content coding were proposed. The main focus was on:
-
4:4:4 8 bit profile defined in RExt
-
New profile HEVC 4:4:4 8 bit screen content coding
National Bodies provided input to profiles related to RExt and Screen Content Coding. Suggestions to create profiles specifically targeted at 4:4:4 screen content with 8 and 10 bits were discussed. Other bit depths are not relevant. There is a relationship to the profiles of the range extensions currently under ballot which allows for intensive discussions. There was no consensus at this meeting to propose a change to the profile specification of the RExt amendment currently under ballot other than adding a 4:4:4 All Intra 16 bit profile. The responses to National Bodies are gathered in N14169 Response to NB on RExt Profiles.
N14165 AHG on coding screen is charged with preparing the evaluation of the CfP.
Dostları ilə paylaş: |