Encoding device, decoding device, encoding method, and decoding method

Abstract

This encoding device comprises a circuit and a memory that is connected to the circuit. In operation, the circuit selects a set of a horizontal conversion base and a vertical conversion base by explicit multiple transform selection (MTS) (S601); uses the selected set to convert a coefficient of a target block (S602); and, when inter-prediction is applied to the target block, derives directionality information using a prediction image applied to the target block and a reconstructed image of a block adjacent to the target block, and, in the selection (S601), uses a table that has been generated by training and that indicates a plurality of candidates for the set to select a set on the basis of the directionality information (S603).

Claims

1. An encoding device comprising: a circuit; and a memory connected to the circuit, wherein the circuit, in operation, selects a pair of horizontal and vertical transformation bases using Explicit MTS (Multiple Transform Selection); transforms the coefficients of a target block using the selected pair; if interpretation is applied to the target block, derives directional information using the predicted image applied to the target block and the reconstructed images of blocks adjacent to the target block; and in the selection, selects the pair based on the directional information using a table generated by learning that shows multiple candidates for the pair.

2. The encoding device according to claim 1, wherein the same table is used in the selection when the inter prediction is applied to the target block and when the intra prediction is applied to the target block.

3. The encoding device according to claim 1, wherein, when the inter prediction is applied to the target block, the selection uses the direction information to select one candidate as the set from the plurality of candidates shown in the table.

4. When the inter prediction is applied to the target block, the encoding device according to claim 1, which selects one of the plurality of candidates and switches, based on the direction information, whether to select the selected candidate as the set or to select a set in which the horizontal transformation basis and the vertical transformation basis of the selected candidate are swapped as the set.

5. The encoding device according to any one of claims 1 to 4, wherein the directional information is derived according to the relationship between (i) the sum of pixel difference values ​​between the predicted image of the upper end of the target block and the reconstructed image of the block adjacent to the upper side of the target block, and (ii) the sum of pixel difference values ​​between the predicted image of the left end of the target block and the reconstructed image of the block adjacent to the left side of the target block.

6. The encoding device according to any one of claims 1 to 4, wherein the directional information is derived according to the relationship between (i) the rate of increase from left to right of the absolute value of the pixel difference between the predicted image of the upper end of the target block and the reconstructed image of the block adjacent to the upper side of the target block, and (ii) the rate of increase from top to bottom of the absolute value of the pixel difference between the predicted image of the left end of the target block and the reconstructed image of the block adjacent to the left side of the target block.

7. A decoding device comprising a circuit and a memory connected to the circuit, wherein the circuit, in operation, selects a pair of horizontal and vertical transformation bases using Explicit MTS (Multiple Transform Selection), inversely transforms the transformed coefficients of a target block using the selected pair, and, if interpretation is applied to the target block, derives directional information using the predicted image applied to the target block and the reconstructed images of blocks adjacent to the target block, and in the selection, selects the pair based on the directional information using a table generated by learning and showing multiple candidates for the pair.

8. The decoding device according to claim 7, wherein the same table is used in the selection when the inter prediction is applied to the target block and when the intra prediction is applied to the target block.

9. The decoding device according to claim 7, wherein, when the inter prediction is applied to the target block, the selection uses the direction information to select one candidate as the set from the plurality of candidates shown in the table.

10. The decoding device according to claim 7, wherein when the inter prediction is applied to the target block, one candidate is selected from the plurality of candidates, and the selected candidate is selected as the set, or a set obtained by swapping the horizontal transformation basis and the vertical transformation basis of the selected candidate is selected as the set, based on the direction information.

11. The decoding device according to any one of claims 7 to 10, wherein the directional information is derived according to the relationship between (i) the sum of pixel difference values ​​between the predicted image of the upper end of the target block and the reconstructed image of the block adjacent to the upper side of the target block, and (ii) the sum of pixel difference values ​​between the predicted image of the left end of the target block and the reconstructed image of the block adjacent to the left side of the target block.

12. The decoding device according to any one of claims 7 to 10, wherein the directional information is derived according to the relationship between (i) the rate of increase from left to right of the absolute value of the pixel difference between the predicted image of the upper end of the target block and the reconstructed image of the block adjacent to the upper side of the target block, and (ii) the rate of increase from top to bottom of the absolute value of the pixel difference between the predicted image of the left end of the target block and the reconstructed image of the block adjacent to the left side of the target block.

13. An encoding method which involves selecting a pair of horizontal and vertical transformation bases using Explicit MTS (Multiple Transform Selection), transforming the coefficients of a target block using the selected pair, deriving directional information using the predicted image applied to the target block and the reconstructed images of blocks adjacent to the target block when interpretation is applied to the target block, and selecting the pair based on the directional information using a table generated by learning that shows multiple candidates for the pair.

14. A decoding method comprising: selecting a pair of horizontal and vertical transformation bases using Explicit MTS (Multiple Transform Selection); inversely transforming the transformed coefficients of a target block using the selected pair; if interpretation is applied to the target block, deriving directional information using the predicted image applied to the target block and the reconstructed images of blocks adjacent to the target block; and in the selection, selecting the pair based on the directional information using a table generated by learning that shows multiple candidates for the pair.

Images