Video compression using boundary based template refinement
Encoding and/or decoding a block of a video frame may be based on a template associated with the block. The length of the template may be determined based on at least one of: a position of the block relative to a boundary, available reference samples above the boundary, and/or the length of a second template of one or more other template derivation algorithms. The length may be adjusted to improve prediction efficiencies and accuracies of template-based predictions.
1 . A method comprising:
determining, by a computing device and based on a quantity of lines of reference samples above a current block of a video, a vertical length of a template above the current block and a reference line for prediction of the template;
determining, based on the reference line determined for prediction of the template, a plurality of predictions of the template respectively associated with a plurality of intra prediction modes;
determining, based on the plurality of predictions of the template, at least one selected intra prediction mode of the plurality of intra prediction modes; and
generating, based on the at least one selected intra prediction mode of the plurality of intra prediction modes, a signal indicating the at least one selected intra prediction mode for an intra prediction associated with the current block.
2 . The method of claim 1 , wherein:
the determining the vertical length of the template and the reference line determined for prediction of the template is for a first intra mode derivation and is further based on at least one of:
a position of the current block relative to a boundary between two coding tree units (CTUs); or
a length of a second template associated with a second intra mode derivation;
a first coding tree unit (CTU), of the two CTUs, comprising the current block; and
the boundary being a top boundary of the first CTU and being a bottom boundary of a second CTU of the two CTUs.
3 . The method of claim 1 , wherein the quantity of line of reference samples above a current block corresponds to a quantity of lines of reconstructed reference samples above the current block that are available for coding the current block, and
wherein at least a portion of the lines of reconstructed reference samples is above a top boundary of a coding tree unit that comprises the current block.
4 . The method of claim 1 , further comprising:
sending, via a bitstream and based on the plurality of predictions of the template, an indication of the at least one selected intra prediction mode of the plurality of intra prediction modes.
5 . The method of claim 1 , further comprising:
encoding or decoding, based on the at least one selected intra prediction mode of the plurality of intra prediction modes and based on reference samples associated with the template, the current block.
6 . The method of claim 1 , wherein the quantity of lines of reference samples above the current block is determined based on a quantity of hardware line buffers configured to store reference lines from an upper coding tree unit (CTU), and wherein a lower CTU comprises the current block and is positioned below the upper CTU.
7 . A method comprising: determining, by a computing device and based on a position of a current block relative to a boundary between two coding tree units (CTUs), a length of a template comprising reconstructed samples above the current block; determining, based on a reference line associated with the template, a plurality of predictions of the template respectively associated with a plurality of intra prediction modes; and sending, via a bitstream and based on the plurality of predictions of the template, an indication of at least one selected intra prediction mode of the plurality of intra prediction modes.
8 . The method of claim 7 , wherein: a first CTU of the two CTUs comprises the current block; and the boundary is a top boundary of the first CTU and is a bottom boundary of a second CTU of the two CTUs.
9 . The method of claim 7 , wherein the determining the length of the template is based on: a top row of the current block being adjacent to the boundary; and a quantity of lines of reconstructed reference samples above the current block that are available for coding the current block.
10 . The method of claim 9 , wherein the length of the template corresponds to: the quantity of lines of reconstructed reference samples above the current block that are available for coding the current block; or one less than the quantity of lines of reconstructed reference samples above the current block that are available for coding the current block.
11 . The method of claim 7 , wherein the determining the length of the template comprises determining, based on a top row of the current block being adjacent to the boundary, a vertical length of the template to be equal to two or zero.
12 . The method of claim 7 , wherein the determining the length of the template comprises determining, based on a top row of the current block being adjacent to the boundary, the length of the template to be equal to a minimum of: one less than a quantity of lines of reconstructed reference samples above the current block that are available for coding the current block; and a value determined based on a length of a left side of the current block.
13 . The method of claim 9 , wherein the quantity of lines of reconstructed reference samples above the current block that are available for coding the current block is determined based on a quantity of line buffers configured to handle data dependencies between the two CTUs in a vertical direction.
14 . The method of claim 7 , wherein the determining the length of the template is based on a top row of the current block not being adjacent to the boundary; and a size of the current block.
15 . A method comprising: determining, by a computing device and based on a position of a current block relative to a boundary between two coding tree units (CTUs), a length of a template comprising reconstructed samples above the current block; determining, based on a reference line associated with the template, a plurality of predictions of the template respectively associated with a plurality of intra prediction modes determining, based on the plurality of predictions of the template, at least one intra prediction mode of the plurality of intra prediction modes; and generating, based on the at least one intra prediction mode, a prediction of the current block.
16 . The method of claim 15 , wherein a first CTU of the two CTUs comprises the current block; and the boundary is a top boundary of the first CTU and is a bottom boundary of a second CTU of the two CTUs.
17 . The method of claim 15 , wherein the determining the length of the template is based on: a top row of the current block being adjacent to the boundary; and a quantity of lines of reconstructed reference samples above the current block that are available for coding the current block.
18 . The method of claim 17 , wherein the length of the template corresponds to: the quantity of lines of reconstructed reference samples above the current block that are available for coding the current block; or one less than the quantity of lines of reconstructed reference samples above the current block that are available for coding the current block.
19 . The method of claim 15 , wherein the determining the length of the template comprises determining, based on a top row of the current block being adjacent to the boundary, a vertical length of the template to be equal to two or zero.
20 . The method of claim 15 , wherein the determining the length of the template comprises determining, based on a top row of the current block being adjacent to the boundary, the length of the template to be equal to a minimum of: one less than a quantity of lines of reconstructed reference samples above the current block that are available for coding the current block; and a value determined based on a length of a left side of the current block.