Add CSP_CENTER. #308
Add CSP_CENTER. #308
Conversation
The chroma_sample_position syntax element is only used for YUV 4:2:0 in the AV1 bitstream. So set chroma_sample_position to ChromaSamplePosition::Unknown for all other YUV formats, including monochrome (YUV 4:0:0). Note: For still images, the most common chroma sample position in practice is the "center" position. Unfortunately the AV1 specification does not have a value for the "center" chroma sample position. See AOMediaCodec/av1-avif#88 and AOMediaCodec/av1-spec#308. This pull request preserves the ChromaSamplePosition::Colocated value that is currently used, but it is likely to be incorrect when YUV 4:2:0 is supported.
07.bitstream.semantics.md
Outdated
| @@ -461,9 +461,11 @@ that subsampling_x is equal to 0 and subsampling_y is equal to 0. | |||
| | 0 | CSP_UNKNOWN | Unknown (in this case the source video transfer function must be signaled outside the AV1 bitstream) | |||
| | 1 | CSP_VERTICAL | Horizontally co-located with (0, 0) luma sample, vertical position in the middle between two luma samples | |||
| | 2 | CSP_COLOCATED | co-located with (0, 0) luma sample | |||
| | 3 | CSP_RESERVED | | |||
| | 3 | CSP_CENTER | In the middle between four luma samples | |||
There was a problem hiding this comment.
I suggest changing the description to "In the center of a 2x2 block of luma samples" or "In the center of each 2x2 block of luma samples"
I also found that ITU-T H.273 (07/21) specifies Chroma420SampleLocType in Section 8.7 Chroma 4:2:0 sample location type. We can also note the correspondence:
CSP_VERTICALis the same asChroma420SampleLocType0CSP_COLOCATEDis the same asChroma420SampleLocType2CSP_CENTERis the same asChroma420SampleLocType1
There was a problem hiding this comment.
I discussed the description of CSP_CENTER with Adrian. We propose the following description, which is similar to the description for CSP_VERTICAL:
Horizontal position in the middle between two luma samples, vertical position in the middle between two luma samples
There was a problem hiding this comment.
Updated to this description. We do have a reference to H.273 in the spec already also, so adding a correspondence might make sense (maybe as a table in a note, rather than extra columns in a table?)
There was a problem hiding this comment.
Yes, the corresponence to Chroma420SampleLocType in H.273 could be added as a table in a note. Note sure if we should aim to minimize changes in the AV1 spec update and refrain from clarification changes.
|
Looks good to me! |
Please let me know if you'd like me to rebuild docs/ in this PR as well.