Joint Collaborative Team on Video Coding (jct-vc) Contribution

8TE planning

8.1General issues for TEs

As a general rule, it was agreed that each TE should be run under the same testing conditions using one software codebase. The general agreed common conditions for experiments were described in the output document JCTVC-B300, as further detailed below.

It is possible to define sub-experiments within particular TEs, eg. TE.xa, TE.xb, etc.

A preliminary TE description is to be approved at the meeting at which the TE plan is established.

A deadline of August 9 was established to express interest in participating in a TE, with finalization of the TE descriptions by August 13 (5 days after the planned availability of TMuC version 0.6).

Experiment descriptions should be written in a way such that it is understood as a JCT-VC output document (written "third party perspective", not a company proponent perspective). The experiment descriptions should generally not express opinions or suggest conclusions – rather, they should just describe what technology will be tested, how it will be tested, who will participate, etc.

Those who proposed technology in the respective context (by this or the previous meeting) can propose a TE or TE sub-experiment. Other subjects would not be designated as TEs.

Any technology must have a cross-check partner to establish a TE – a single proponent is not enough. It is highly desirable have more than just one proponent and one cross-checker.

It is strongly recommended to plan resources carefully and not waste time on technology that may have little benefit – it is also within the responsibility of the TE coordinator to take care of this.

Each participant is required to submit a report – measures will be taken if this principle is violated without good reason.

A summary report by the coordinator is expected to be provided to the subsequent meeting.

Particular TEs planned at this meeting are summarized in the following subsections of this report.

A particularly important TE is TE12, for testing of the individual coding tools in the TMuC.

The TE description for each planned TE is described in an associated output document JCTVC-B3xx for TExx, where "xx" is the TE number.

Good results from a TE do not impose an obligation on the group to accept the result (e.g., if the expert judgment of the group is that further data is needed or that the test methodology was flawed).

8.2Common conditions for testing

Common conditions for experiment testing were selected by the group and described in output document JCTVC-B300.

8.3TE1 Decoder-side Motion Vector Derivation

Continuation of a similar prior TE, coordinated by Mathias Wien (RWTH), summarized as follows:

• 
  Implementation in TMuC
  
• 
  6 proposals included
  

• 
  2 sub-experiments: Inter prediction, direct mode
  
• 
  Try to explore interdependency with TMuC tools
  
• 
  Only high complexity configuration
  

8.4TE2 IBDI and memory compression (old TE2)

Continuation of the prior TE on memory compression, coordinated by Takeshi Chujoh (Toshiba), with additional emphasis on:

• 
  Sizes of access units (should be comparable, and idea about the overhead that comes due to overlaps in random access)
  
• 
  In which parts of the codec is IBDI necessary – only for processing, or also for storage?
  
• 
  Due to the need for fixed bit rate per access unit, subjective impairments may appear locally
  -> subjective investigation needed.
  
• 
  Complexity investigation of different methods
  
• 
  Is built-in spatial scalability useful e.g. for power saving?
  

Some discussed aspects:

• 
  Implementation in TMuC
  
• 
  Use BD rate/PSNR computation interpolation mode 3 (default).
  
• 
  Also measure the distortion per access block (in order to identify whether there are unusually high distortions locally).
  

An AHG will study related issues beyond the scope of the TE (e.g. relation with MC etc.).

8.5TE3 Inter Prediction

This is a continuation of a related prior TE, to be coordinated by Andreas Krutz (TUB).

Sub-experiments will include:

• 
  Warped MC and second order prediction
  
• 
  Flexible motion partitioning (relation with TE12)
  
• 
  Multi-hypothesis inter prediction (investigate potentially more efficient methods than with current B picture mechanisms)
  
• 
  Improved inter prediction with enhanced MC filter (relation with TE12)
  

8.6TE4 Variable length coding

This TE will be coordinated by X. Wang (Qualcomm). It is focused on JCTVC-B098 (Qualcomm, Nokia, Tandberg) CBF coding, new interleaving method in large transform blocks

This work needs to wait for new implementation of the VLC coding in the TMuC software.

This is a very focused experiment on potentially improving the VLC entropy coding in the TMuC.

8.7TE5 TMuC Intra prediction simplification

This TE will be coordinated by T.K. Tan (NTT Docomo). It is focused on investigation of JCTVC-B093 and JCTVC-B118 in a TE (proposing simplifications to be verified for potential adoption by the next meeting).

A question that was raised (not investigated by this TE) was whether the benefit of the "harmonized design" which has 17 prediction directions in case of 4x4 will be tested. It was confirmed that this will be done in TE12 (17 versus 9 directions).

8.8TE6 Intra prediction improvements

This TE will be coordinated by Ali Tabatabai (Sony). It will include investigation of the following:

• 
  JCTVC-B040 and JCTVC-B042 (bidirectional part) – investigated in a common TMuC environment
  
• 
  JCTVC-B109 in relation to the method of prediction direction prediction that is already in the TMuC
  
• 
  JCTVC-B112
  
• 
  JCTVC-A109 (Mediatek) and JCTVC-A118 (Microsoft)
  

Tests will be conducted in both the low and high complexity TMuC configurations.