Flipped templates for reconstruction-reordered intra block copy
A current block may be encoded/decoded based on a reference block flipped relative to the current block. An indication of whether a value of a parameter (e.g., a block vector difference) indicating the reference block matches a value of a predictor for the parameter (e.g., a block vector difference predictor) may be entropy encoded. The predictor may be selected from among a plurality of candidates based on costs associated with the plurality of candidates associated with candidate reference blocks. The cost of each candidate may be based on comparisons between a template of a current block and templates of the candidate reference blocks, where the templates of the candidate reference blocks have a shape flipped relative to that of the current block.
1 . A method comprising:
determining, by a computing device, a plurality of candidate templates, for a plurality of reference block (RB) candidates associated with a plurality of block vector differences (BVD) candidates, that are flipped in a direction relative to a current template for a current block (CB) of content;
based on comparisons between the current template and each candidate template of the plurality of candidate templates, determining a plurality of template matching costs associated with the plurality of BVD candidates;
based on the plurality of template matching costs, selecting a BVD candidate, of the plurality of BVD candidates, as a BVD predictor;
decoding an indication of whether a value of a symbol of a BVD matches a value of a symbol of the BVD predictor; and
based on the indication and a value of the symbol of the BVD predictor, determining a value of the symbol of the BVD.
2 . The method of claim 1 , wherein each candidate template has a shape, of the current template, flipped in the direction.
3 . The method of claim 1 , wherein the comparisons between the current template and each candidate template are based on comparisons between samples in the candidate template and corresponding samples in the current template, wherein the samples in the candidate template have locations, relative to associated RB candidates, that are flipped relative to locations of the corresponding samples, relative to the CB.
4 . The method of claim 1 , wherein each candidate template, corresponding to a reference block candidate, comprises:
based on the direction being horizontal, a same quantity of columns of samples to the right of the RB candidate as a quantity of columns of samples, in the current template, to the left of the CB; or
based on the direction being vertical, a same quantity of rows of samples below the RB candidate as a quantity of rows of samples, in the current template, above the CB.
5 . The method of claim 3 , wherein the comparisons between the current template and each candidate template do not include comparisons between samples, in the candidate template that overlap with the current template, and corresponding samples in the current template.
6 . The method of claim 1 , further comprising:
decoding a codeword comprising a subset of symbols indicating the BVD, wherein the subset of symbols exclude the symbol of the BVD; and
determining the plurality of BVD candidates based on the codeword.
7 . The method of claim 1 , wherein the symbol of the BVD is one or more of:
a sign symbol of the BVD, or
a magnitude symbol of the BVD.
8 . The method of claim 1 , wherein the BVD candidate, of the plurality of BVD candidates, is selected as the BVD predictor based on being associated with a smallest template matching cost of the plurality of template matching costs.
9 . The method of claim 1 , further comprising:
determining a RB based on the BVD, wherein the CB or the RB is flipped in the direction relative to the other.
10 . A method comprising:
determining, by a computing device, a block vector difference (BVD) based on a difference between a block vector (BV), of a current block (CB) of content, and a block vector predictor (BVP);
determining a plurality of candidate templates, for a plurality of reference block (RB) candidates associated with a plurality of BVD candidates, that are flipped in a direction relative to a current template for the CB;
based on comparisons between the current template and each of the candidate templates of the RB candidates associated with the BVD candidates, determining a plurality of template matching costs for the plurality of BVD candidates;
based on the plurality of template matching costs, selecting a BVD candidate, of the BVD candidates, as a BVD predictor; and
encoding an indication of whether a value of a symbol of the BVD matches a value of a symbol of the BVD predictor.
11 . The method of claim 10 , wherein each candidate template has a shape, of the current template, flipped in the direction.
12 . The method of claim 10 , wherein the determining the plurality of template matching costs is based on, for each candidate template of the plurality of candidate templates, comparisons between samples in the candidate template and corresponding samples in the current template, wherein the samples in the candidate template are at locations, relative to a corresponding RB candidate, corresponding to locations, of the corresponding samples in the current template relative to the CB, flipped in the direction.
13 . The method of claim 10 , wherein the RB candidates are displaced relative to the CB by a sum of the respective BVD candidates and the BVP.
14 . The method of claim 10 , wherein:
the BVD candidates comprise a first BVD candidate and a second BVD candidate; and
a value of the symbol of the first BVD candidate is different from a value of the symbol of the second BVD candidate, wherein the first BVD candidate is selected as the BVD predictor.
15 . The method of claim 10 , wherein the encoding the indication further comprises arithmetically encoding the indication based on a probability model indicating:
a probability of a least probable symbol for the indication; and
a value of a most probable symbol for the indication.
16 . The method of claim 10 , wherein the RB candidates are located in a reference region corresponding to the direction.
17 . The method of claim 10 , wherein the CB is coded in a reconstruction-reordered intra block copy mode.
18 . A method comprising:
determining, by a computing device, a plurality of candidate templates, for a plurality of reference block (RB) candidates associated with a plurality of block vector differences (BVD) candidates, and based on a current template for a current block (CB) of content,
wherein each candidate template, of the plurality of candidate templates, has a shape, of the current template, flipped in a direction;
based on comparisons between the current template and each candidate template of the plurality of candidate templates, determining a plurality of template matching costs corresponding to the plurality of BVD candidates;
based on the plurality of template matching costs, selecting a BVD candidate, of the plurality of BVD candidates, as a BVD predictor;
decoding information based on a BVD and the BVD predictor; and
determining, based on the decoded information, a value of a symbol of the BVD.
19 . The method of claim 18 , wherein the decoded information comprises an indication of whether a value of a symbol of the BVD matches a value of a corresponding symbol of the BVD predictor.
20 . The method of claim 18 , wherein the current template comprises:
a first portion comprising a first quantity of rows of samples above the CB; and
a second portion comprising a second quantity of columns of samples to the left of the CB, and
wherein each candidate template, of the plurality of candidate templates, corresponding to a RB candidate of the plurality of RB candidates, comprises:
based on the direction being a horizontal direction:
a first portion comprising the first quantity of rows of samples above the RB candidate; and
a second portion comprising the second quantity of columns of samples to the right of the RB candidate, or
based on the direction being a vertical direction:
a first portion comprising the first quantity of rows of samples below the RB candidate; and
a second portion comprising the second quantity of columns of samples to the left of the RB candidate.