Method, device, and recording medium storing bit stream, for encoding/decoding image
The invention relates to a method for encoding/decoding an image. The image decoding method according to the invention comprises the steps of: obtaining diagonal partition information on a current block; determining a diagonal partition structure of the current block using the diagonal partition information; and diagonally partitioning the current block into a first and a second area based on the determined diagonal partition structure, the current block being a leaf node of a square or rectangular partition.
1. A method for decoding an image comprising:
dividing, based on geometric motion partition, a current block into a first partition and a second partition;
obtaining a first prediction block of the current block based on motion information of the first partition;
obtaining a second prediction block of the current block based on motion information of the second partition; and
obtaining a final prediction block of the current block by performing a weighted sum of the first prediction block and the second prediction block,
wherein the geometric motion partition is performed based on division information indicating a direction of a division line dividing the current block,
wherein a shape of at least one of the first partition and the second partition is triangular, and
wherein the geometric motion partition is applied to the current block only when the current block does not perform sub-block-based inter prediction.
2. The method of claim 1 , wherein the sub-block-based inter prediction includes sub-block-based temporal candidate prediction.
3. The method of claim 1 , wherein the motion information of the first partition and the second partition is derived based on a merge mode, respectively.
4. The method of claim 3 , wherein the first partition has a merge index different from the second partition, and
wherein the merge index indicates a merge candidate of the first partition.
5. The method of claim 1 , wherein the geometric motion partition is performed only for a coding block that is no longer divided into smaller coding blocks through tree-based block division, and
wherein the tree-based block division includes at least one of quad-tree division or binary-tree division.
6. The method of claim 1 , wherein the weighted sum of the first prediction block and the second prediction block is performed based on a predetermined weight set, and
wherein the predetermined weight set is determined, based on a position of a current sample in the current block, as one of a plurality of weight set candidates.
7. The method of claim 6 , wherein each of the plurality of weight set candidates includes a first weight to be applied to a first prediction sample belonging to the first prediction block and a second weight to be applied to a second prediction sample belonging to the second prediction block, and
wherein a sum of the first weight and the second weight is 8.
8. A method for encoding an image, wherein a current block in the image is divided, based on geometric motion partition, into a first partition and a second partition, the method comprising:
obtaining a first prediction block corresponding to the first partition in the current block;
obtaining a second prediction block corresponding to the second partition in the current block;
obtaining a final prediction block of the current block by performing a weighted sum of the first prediction block and the second prediction block; and
generating a bitstream by encoding an original block of the current block based on the final prediction block,
wherein the bitstream includes motion information indicating the first prediction block of the first partition, motion information indicating the second prediction block of the second partition, and division information indicating a division direction of the geometric motion partition,
wherein a shape of at least one of the first partition and the second partition is triangular, and
wherein the geometric motion partition is applied to the current block only when the current block does not perform sub-block-based inter prediction.
9. A non-transitory computer-readable medium for storing data associated with an image signal, comprising:
a data stream encoded by an encoding method, wherein a current block in the image signal is divided, based on geometric motion partition, into a first partition and a second partition, and
wherein the encoding method comprises:
obtaining a first prediction block corresponding to the first partition in the current block;
obtaining a second prediction block corresponding to the second partition in the current block; and
obtaining a final prediction block of the current block by performing a weighted sum of the first prediction block and the second prediction block; and
generating a bitstream by encoding an original block of the current block based on the final prediction block,
wherein the bitstream includes motion information indicating the first prediction block of the first partition, motion information indicating the second prediction block of the second partition, and division information indicating a division direction of the geometric motion partition,
wherein a shape of at least one of the first partition and the second partition is triangular, and
wherein the geometric motion partition is applied to the current block only when the current block does not perform sub-block-based inter prediction.