Affine model-based image encoding/decoding method and device
In an image encoding/decoding method and device according to the present invention, a candidate list for motion information prediction of a current block is generated, a control point vector of the current block is derived on the basis of the candidate list and a candidate index, a motion vector of the current block is derived on the basis of the control point vector of the current block, and inter-prediction with respect to the current block can be performed by means of the motion vector.
1. A method of decoding an image, comprising:
generating a candidate list including merge candidates for motion information prediction of a current block in the image, wherein the merge candidates include at least one of a plurality of affine candidates or a subblock-based temporal candidate, the plurality of affine candidates includes a spatial candidate derived from a neighboring block adjacent to the current block, and the neighboring block includes at least one of a left neighboring block or a top neighboring block;
deriving, in units of subblocks of the current block, a motion vector of the current block based on the candidate list and a candidate index, the candidate index specifying one of the merge candidates in the candidate list;
generating a prediction block of the current block by performing inter prediction for the current block based on the motion vector; and
reconstructing the current block based on the prediction block,
wherein, in response to a case where a merge candidate specified by the candidate index is one of the plurality of the affine candidates, deriving the motion vector comprises:
deriving a control point vector of the current block based on the specified merge candidate; and
deriving the motion vector based on the control point vector of the current block, and
wherein, in response to a case where the specified merge candidate is the subblock-based temporal candidate, a motion vector of each subblock belonging to the current block is derived based on a motion vector of a subblock in a collocated block corresponding to the each subblock in the current block.
2. The method of claim 1 , wherein the control point vector of the current block includes at least one of a first control point vector corresponding to a top left sample of the current block, a second control point vector corresponding to a top right sample of the current block, or a third control point vector corresponding to a bottom left sample of the current block.
3. The method of claim 2 , wherein a control point vector of the spatial candidate is derived by considering whether a boundary of the current block is located on a boundary of coding tree block (CTU).
4. The method of claim 3 , wherein, in response to a case where the boundary of the current block is not located on the boundary of the CTU, the control point vector of the spatial candidate is derived based on a control point vector of the neighboring block, and
wherein, in response to a case where the boundary of the current block is located on the boundary of the CTU, the control point vector of the spatial candidate is derived based on a motion vector of the neighboring block.
5. The method of claim 2 , wherein, in response to the case where the specified merge candidate is the one of the plurality of the affine candidates, the motion vector of the current block is derived based on at least one of the first control point vector, the second control point vector, a position of the subblock, or a size of the current block.
6. The method of claim 1 , wherein the collocated block belongs to a picture different from the current block, and
wherein the collocated block is representative of a block at a position shifted by a temporal vector from a position of the current block.
7. The method of claim 6 , wherein the temporal vector is determined based on the left neighboring block among neighboring blocks adjacent to the current block.
8. A method of encoding an image, comprising:
generating a candidate list including merge candidates for motion information prediction of a current block in the image, wherein the merge candidates include at least one of a plurality of affine candidates or a subblock-based temporal candidate, the plurality of affine candidates includes a spatial candidate derived from a neighboring block adjacent to the current block, and the neighboring block includes at least one of a left neighboring block or a top neighboring block;
deriving, in units of subblocks of the current block, a motion vector of the current block based on the candidate list;
generating a prediction block of the current block by performing inter prediction for the current block based on the motion vector; and
generating a residual block of the current block based on the prediction block,
wherein a candidate index specifying one of the merge candidates in the candidate list is encoded into a bitstream,
wherein, in response to a case where the motion vector of the current block is derived based on one of the plurality of the affine candidates, deriving the motion vector comprises:
deriving a control point vector of the current block based on the one of the plurality of the affine candidates; and
deriving the motion vector based on the control point vector of the current block, and
wherein, in response to a case where the motion vector of the current block is derived based on the subblock-based temporal candidate, a motion vector of each subblock belonging to the current block is derived based on a motion vector of a subblock in a collocated block corresponding to the each subblock in the current block.
9. A method of transmitting a bitstream, the method comprising:
generating a candidate list including merge candidates for motion information prediction of a current block, wherein the merge candidates include at least one of a plurality of affine candidates or a subblock-based temporal candidate, the plurality of affine candidates includes a spatial candidate derived from a neighboring block adjacent to the current block, and the neighboring block includes at least one of a left neighboring block or a top neighboring block;
deriving, in units of subblocks of the current block, a motion vector of the current block based on the candidate list;
generating a prediction block of the current block by performing inter prediction for the current block based on the motion vector;
generating a residual block of the current block based on the prediction block;
encoding the residual block to generate the bitstream; and
transmitting the bitstream,
wherein a candidate index specifying one of the merge candidates in the candidate list is encoded into the bitstream,
wherein, in response to a case where the motion vector of the current block is derived based on one of the plurality of the affine candidates, deriving the motion vector comprises:
deriving a control point vector of the current block based on the one of the plurality of the affine candidates; and
deriving the motion vector based on the control point vector of the current block, and
wherein, in response to a case where the motion vector of the current block is derived based on the subblock-based temporal candidate, a motion vector of each subblock belonging to the current block is derived based on a motion vector of a subblock in a collocated block corresponding to the each subblock in the current block.