Method and apparatus for decoding/encoding video signal using transform derived from graph template
A method for decoding a video signal using a graph-based transform, where the method includes: receiving a template index for a target unit from the video signal, wherein the template index indicates a graph-based transform template to be applied to the target unit; deriving a graph-based transform matrix for the target unit corresponding to the template index; and decoding the target unit based on the graph-based transform matrix. In some scenarios, this enables efficient compression of inter-predicted residual blocks by using graph-based transform templates. The graph-based transform template may enable adaptive transform selection, thus achieving improved coding gain. Further, by defining a graph-based transform template or a template index, coding with a low-complexity is enabled that can reduce the number of bits to be transmitted to the decoder.
1. A method for decoding a video signal using a graph-based transform, the method comprising:
receiving a template index for a target unit from the video signal, wherein the template index indicates a graph-based transform template to be applied to the target unit;
deriving a graph-based transform matrix for the target unit corresponding to the template index; and
decoding the target unit based on the graph-based transform matrix,
wherein the template index corresponds to one template graph in a predetermined template graph set,
wherein the predetermined template graph set includes a plurality of template graphs, and
wherein the plurality of template graphs have a pattern in which a corner region in the target unit has a smaller edge weight than other regions in the target unit.
2. The method of claim 1 ,
wherein the predetermined template graph set is determined based on a specific pattern of the target unit, and
wherein the specific pattern includes at least one of an edge weight pattern, a pixel value pattern, and a node connection pattern.
3. A method for encoding a video signal using a graph-based transform, the method comprising:
extracting a graph parameter, wherein the graph parameter includes at least one of a vertex parameter set and an edge parameter set;
generating a base template graph based on the graph parameter;
generating a template graph set based on adjustment of an edge weight of the base template graph;
determining a template graph corresponding to a target unit from the template graph set;
acquiring a graph-based transform kernel through a spectral decomposition of the template graph; and
generating a template index corresponding to each of a plurality of template graphs included in the template graph set,
wherein the template graph set includes a plurality of template graphs, and
wherein the plurality of template graphs have a pattern in which a corner region in the target unit has a smaller edge weight than other regions in the target unit.
4. The method of claim 3 ,
wherein the template graph set is generated based on a specific pattern of the target unit, and
wherein the specific pattern includes at least one of an edge weight pattern, a pixel value pattern, and a node connection pattern.
5. A device for decoding a video signal using a graph-based transform, the device comprising:
an entropy-decoding unit configured to receive a template index for a target unit from the video signal; and
an inverse transform unit configured to derive a graph-based transform matrix for the target unit corresponding to the template index, and to perform an inverse-transform the target unit based on the graph-based transform matrix,
wherein the template index indicates a graph-based transform template to be applied to the target unit,
wherein the template index corresponds to one template graph in a predetermined template graph set,
wherein the predetermined template graph set includes a plurality of template graphs, and
wherein the plurality of template graphs have a pattern in which a corner region in the target unit has a smaller edge weight than other regions in the target unit.
6. The device of claim 5 ,
wherein the predetermined template graph set is determined based on a specific pattern of the target unit, and
wherein the specific pattern includes at least one of an edge weight pattern, a pixel value pattern, and a node connection pattern.
7. A device for encoding a video signal using a graph-based transform, the device comprising:
a graph signal generation unit configured to
generate a base template graph based on a graph parameter, wherein the graph parameter includes at least one of a vertex parameter set and an edge parameter set, and
generate a template graph set based on adjustment of an edge weight of the base template graph;
a graph-based transform template determination unit configured to determine a template graph corresponding to a target unit from the template graph set;
a graph-based transform acquisition unit configured to acquire a graph-based transform kernel through a spectral decomposition of the determined template graph; and
a template index generation unit configured to generate a template index corresponding to each of a plurality of template graphs included in the template graph set,
wherein the template graph set includes a plurality of template graphs, and
wherein the plurality of template graphs have a pattern in which a corner region in the target unit has a smaller edge weight than other regions in the target unit.
8. The device of claim 7 ,
wherein the template graph set is generated based on a specific pattern of the target unit, and
wherein the specific pattern includes at least one of an edge weight pattern, a pixel value pattern, and a node connection pattern.