JVET-AK0179 AHG9: On Lens Optical Correction SEI message [J. Samuelsson-Allendes, S. Deshpande (Sharp)]
This input contribution relates to the Lens Optical Correction (LOC) SEI message which is included in the technologies under consideration document for future versions of VSEI. The proposals in this document are intended to broaden the applicability of the LOC SEI message and are asserted to provide clarifications of the usage of the LOC SEI message. Specifically, the following items are proposed:
- Separate the case of “no information” from “no distortion”
- Remove two sentences that include the term “by default”
- Express focal location as ratio of picture width and height, respectively
The existing semantics for the presence flag currently indicate no expected distortion, when in general they simply indicate that no information is provided.
It was questioned if there are use cases where parameters in some category would be available when other categories are not available.
A comment was that most real lenses have some distortion in all three categories.
A question was raised about if computer generated content such as generative AI may wish to emulate lens distortion in some but not all categories.
There was agreement to Item 1.
There was agreement to Item 2.
Regarding Item 3, it would be useful to have information on how focal parameters have been signalled in other standards by JVET or MPEG.
It is noted that the semantics language for loc_focal_center_x and _y use “default”, when it would be more appropriate to have inference language.
It is suggested to check if constraints on value ranges are necessary.
Further discussion requested on Item 3.
Decision: Adopt Items 1 and Item 2 (editorial).
Further discussed 1910 on 19 Jan.
The -v2 version adds the following to its abstract:
4. Add missing inferred values and value ranges for several syntax elements
The second version of the proposal adds an option B for item 3, where the focal location is expressed in luma samples. A fourth item is also added to define inferred values and value ranges for several syntax elements.
Decision: Adopt Item 3 option A and Item 4.
Follow-up discussion Monday 20 Jan. 1005 (chaired by JRO). It was commented that there is an ambiguity in the definition of the syntax elements loc_rd_poly*, some of which are u(25), some (i25), some unclarity about correctness and necessity of the ranges, and due to the fact that a division occurs, it is unclear what the result would be, in which floating point format it comes.
Decision: It was agreed to modify the decisions above by removing the ranges in item 4, and conduct further study about correctness of syntax and semantics, and the necessity to impose range constraints.