Opened 5 years ago
Closed 5 years ago
#258 closed defect (wontfix)
abs_delta_poc_st should be abs_delta_poc_st_minus1
Reported by: | Vadim | Owned by: | |
---|---|---|---|
Priority: | minor | Milestone: | |
Component: | spec | Version: | |
Keywords: | Cc: | ksuehring, bbross, XiangLi, fbossen, jvet@… |
Description
abs_delta_poc_st[ listIdx ][ rplsIdx ][ i ], when the i-th entry is the first STRP entry in ref_pic_list_struct( listIdx, rplsIdx ) syntax structure, specifies the absolute difference between the picture order count values of the current picture and the picture referred to by the i-th entry, ... The value of abs_delta_poc_st[ listIdx ][ rplsIdx ][ i ] shall be in the range of 0 to 215 − 1, inclusive.
DeltaPocSt[ listIdx ][ rplsIdx ][ i ] = ( strp_entry_sign_flag[ listIdx ][ rplsIdx ][ i ]) ? abs_delta_poc_st[ listIdx ][ rplsIdx ][ i ] : 0 − abs_delta_poc_st[ listIdx ][ rplsIdx ][ i ]
RefPicPocList[ i ][ j ] = pocBase − DeltaPocSt[ i ][ RplsIdx[ i ] ][ j ]
In the current draft, DeltaPocSt equal 0 is possible, so the reference picture with the POC equal to the current picture POC value will be added to the reference picture list and for other entries in the ref_pic_list_struct.
abs_delta_poc_st_minus1 should be signaled instead and
DeltaPocSt[ listIdx ][ rplsIdx ][ i ] = ( strp_entry_sign_flag[ listIdx ][ rplsIdx ][ i ] ? 1 + abs_delta_poc_st_minus1[ listIdx ][ rplsIdx ][ i ] : -1 − abs_delta_poc_st_minus1[ listIdx ][ rplsIdx ][ i ] )
Change history (2)
comment:1 Changed 5 years ago by yk
comment:2 Changed 5 years ago by bbross
- Resolution set to wontfix
- Status changed from new to closed
Inclusion of the current picture into an RPL entry has been disabled by a constraint: "The picture referred to by each entry in RefPicList[ 0 ] or RefPicList[ 1 ] shall not be the current picture." Maybe it's better to change "each" in the sentence to "any".
Allowing a zero-valued POC delta is still needed to enable a particular reference picture to be repeated in consecutive entries in an RPL. Allowing repeating a particular reference picture multiple times in an RPL enables assigning different weights for that reference picture in weighted prediction.
Therefore, the suggested change here should not be made, and this ticket can be closed.