International organisation for standardisation organisation internationale de normalisation

5.6.5Interlace

5.6.5.1.1.1.1.1.1JCTVC-H0056 AHG13: On interlace indication [K. Chono, H. Aoki (NEC)]

This contribution was summarized in the AHG13 report as:

• 
  Move the VUI flag in the AHG13 starting text to SPS and modify its definition to specify the type of the associated coded sequence in addition to the presence of the SEI message in each of access units.
  
• 
  One-bit interlace field indication in SEI message to indicate the parity of each field.
  
• 
  Modified access unit delimiter RBSP to indicate dynamically changing source material properties.
  

The first element proposes two aspects: 1) moving the location of the SPS-level flag, and 2) defining the flag as an indication that the coded video sequence is interlaced fields.

The second element proposes to reduce the length of the field in the SEI message to use only 1 bit rather than 4 bits.

The last element was provided as information, rather than for immediate adoption.

5.6.5.1.1.1.1.1.2JCTVC-H0588 AHG13: Proposed interlace format indication for HEVC [K. Sugimoto, S. Sekiguchi (Mitsubishi)] [late]

This contribution was summarized in the AHG13 report as:

• 
  Define a constraint that a coded sequence shall contain coded pictures consisting of frames only (progressive or interlace) or fields only.
  
• 
  One-bit interlace format indication in SPS or VUI
  
• 
  One-bit interlace field indication in SEI message to indicate the parity of each field.
  

This is very similar to JCTVC-H0056.

It was remarked that the question of whether the content is coded as fields is different than the question of whether the input sampling structure was progressive or interlaced.

It was suggested to avoid the use of interlaced sampling as a concept in the specification – rather, simply using the geometric concept of top and bottom fields (although a note with some acknowledgement of a relationship may be appropriate).

It was suggested that it may be desirable, when frames are coded, to enable an indication that the frames are interlaced and which field is temporally first.

5.6.5.1.1.1.1.1.3JCTVC-H0666 Interlace Indicators [Chad Fogg (Harmonic)] [late]

This contribution was summarized in the AHG13 report as:

• 
  A flag to indicate the scanning method of the original source of a coded sequence.
  
• 
  A flag to indicate whether a coded sequence contains frames only or fields only.
  
• 
  A flag to indicate the parity of each field.
  
• 
  A flag to indicate whether the coded frame or field is a "dummy frame" resulting from a repetition process, such as 3:2 pulldown.
  

One thing not in the proposal was the aforementioned indication which field is temporally first when a picture is a frame. A suggestion was to provide a flag to indicate a conceptual pairing of fields – i.e. which field is first in a pair.

The three remaining bits of the byte would be reserved.

A BoG (coordinated by K. Chono & C. Fogg) was asked to prepare text. The result was reported in JCTVC-H0720.

5.6.5.1.1.1.1.1.4JCTVC-H0222 Introduction of AVC derivation process of chroma motion vector for interlace format support [K. Kazui, J. Koyama, S. Shimada, A. Nakagawa (Fujitsu)]

This contribution was summarized in the AHG13 report as:

• 
  SPS flag to specify the scanning method of an input source.
  
• 
  One-bit interlace field indication in slice header when SPS flag indicates that the input source is interlaced scan.
  
• 
  AVC-style chroma MV phase adjustment.
  

It proposes part of the prior JCTVC-G196: field motion compensation for chroma as in AVC.

The claimed improvement was 0.4 dB average BD PSNR benefit (up to 1 dB) for chroma components (using self-selected interlaced content). No cross-verification was provided.

It was asked whether the improvement is visible. It was remarked by a non-proponent that there previously has been a conclusion that there is some visual benefit in AVC chroma artefact reduction as a result of this technique.

If adopted, we might need to have a profile that excludes it. It was commented that if we are willing to customize the decoding process for field coding, we should probably not just do this, but should go further – however, we really don't plan that. No action taken.

5.6.5.1.1.1.1.1.5JCTVC-H0720 BoG report on interlace indicators [K. Chono, C. Fogg, D. Flynn, S. Sekiguchi]

After further review and discussion, it was agreed to have a flag that indicates the mandatory presence of an SEI message that indicates whether all pictures are frames vs. all pictures being fields, and conveys four picture-level property flags as described above.

The text was subsequently reviewed and refined by the group, with the result uploaded as the -v4 version.

Decision: Adopt as reviewed and uploaded as -v4 of H0720.