Method and apparatus for last coefficient coding for adaptive transform based video compression
Coding of the last coded coefficient position is performed by basing the coding of the y coordinate of the position of the last coded coefficient on knowledge of the size of the partial transform used to generate a block of coefficients from a block of video pixels. This enables a context adaptive coding of the last coded coefficient parameter to be performed much more efficiently.
1 . A method for encoding a block of transformed coefficients, comprising:
transforming a block of image values using a partial transform to obtain a block of transformed coefficients such that positions of a subset of the transformed coefficients of the block of transformed coefficients equal to zero are known; and
entropy coding information representing a position in the block of transformed coefficients of a last non-zero transformed coefficient,
wherein the entropy coding of information representing the position depends on a size of the partial transform.
2 . The method of claim 1 , wherein an information representing the size of the partial transform is encoded in video data.
3 . The method of claim 1 ,
wherein, the information representing the position represents two coordinates of the last non-zero transformed coefficient, the two coordinates being split into a prefix and a suffix; and
wherein the entropy coding of the information representing the position comprises entropy encoding the prefix of at least one of the two coordinates based on the size of the partial transform.
4 . The method of claim 1 , wherein an adaptive set of partial transforms is used to transform the block of image values.
5 . The method of claim 4 , wherein the adaptive set of partial transforms is learned applying an on-the-fly block based transform learning scheme.
6 . The method of claim 5 , wherein the on-the-fly block based transform learning scheme comprises a classification of residual blocks of a picture in a plurality of classes and a generation of an adapted set of transforms in which a transform is determined for each class of the plurality using a minimization of a reconstruction error for that class.
7 . The method of claim 6 , wherein the classification is based on a singular value decomposition and Karhunen-Loève Transforms are used to generate the adapted set of transforms.
8 . The method of claim 1 , wherein a content-adaptive method is used to determine the size of the partial transform.
9 . An apparatus for encoding a block of transformed coefficients, comprising:
transforming a block of image values using a partial transform to obtain a block of transformed coefficients such that positions of a subset of the transformed coefficients of the block of transformed coefficients equal to zero are known;
entropy coding information representing a position in the block of transformed coefficients of a last non-zero transformed coefficient,
wherein the entropy coding of information representing the position depends on a size of the partial transform.
10 . The apparatus of claim 9 , wherein an information representing the size of the partial transform is encoded in a bitstream.
11 . The apparatus of claim 9 ,
wherein the information representing the position represents two coordinates of the last non-zero transformed coefficient, the two coordinates being split into a prefix and a suffix; and
wherein the entropy coding of the information representing the position comprises entropy encoding the prefix of at least one of the two coordinates based on the size of the partial transform.
12 . A method for decoding a block of transformed coefficients from video data, comprising:
entropy decoding a code to generate a position of a last coded coefficient,
wherein the entropy decoding of the code depends on a size of a partial transform; and
inverse transforming the block of transformed coefficients using the partial transform to obtain a block of image values, the size of the partial transform allowing deriving positions of transform coefficients of the block of transformed coefficients equal to zero.
13 . The method of claim 12 , wherein an information representing the size of the partial transform is decoded from video data.
14 . The method of claim 12 ,
wherein the code represents two coordinates of the last non-zero transformed coefficient, the two coordinates being split into a prefix and a suffix; and
wherein the entropy decoding of the code comprises entropy decoding the prefix of at least one of the two coordinates based on the size of the partial transform.
15 . The method of claim 12 , wherein the partial transform is signaled in the video data and belong to a set of partial transforms determined for an image comprising the block of image values.
16 . The method of claim 15 wherein, the set of partial transforms comprises Karhunen-Loève Transforms.
17 . An apparatus for decoding a block of transformed coefficients from video data, comprising:
entropy decoding a code to generate a position of a last coded coefficient,
wherein the entropy decoding of the code depends on a size of a partial transform; and
inverse transforming the block of transformed coefficients using the partial transform to obtain a block of image values, the size of the partial transform allowing deriving positions of transform coefficients of the block of transformed coefficients equal to zero.
18 . The apparatus of claim 17 , wherein an information representing the size of the partial transform is decoded from video data.
19 . The apparatus of claim 17 ,
wherein the code represents two coordinates of the last non-zero transformed coefficient, the two coordinates being split into a prefix and a suffix; and
wherein the entropy decoding of the code comprises entropy decoding the prefix of at least one of the two coordinates based on the size of the partial transform.
20 . A non-transitory computer-readable storage medium having stored thereon instructions that, when executed, decode a set of transformed coefficients by:
entropy decoding a code to generate a position of a last coded coefficient,
wherein the entropy decoding of the code depends on a size of a partial transform; and
inverse transforming the block of transformed coefficients using the partial transform to obtain a block of image values, the size of the partial transform allowing deriving positions of transform coefficients of the block of transformed coefficients equal to zero.