Inter prediction-based video encoding and decoding
A video decoding method is disclosed. The method includes: determining a partitioning structure of the target block; reconstructing motion vectors, which correspond to a plurality of partitions split from the target block by the partitioning structure of the target block; using the motion vectors to generate prediction blocks that respectively correspond to the partitions and are equal in size to the target block; generating a prediction block of the target block by weighted averaging the prediction blocks corresponding to the partitions; and reconstructing the target block based on the prediction block of the target block.
1 . A video decoding method of decoding a target block, the method comprising:
determining a partitioning parameter for the target block;
reconstructing motion vectors, which correspond to a plurality of partitions split from the target block by the partitioning parameter of the target block;
using the motion vectors to generate prediction blocks that respectively correspond to the partitions and are equal in size to the target block;
generating a prediction block of the target block by weighted averaging the prediction blocks corresponding to the partitions; and
reconstructing the target block based on the prediction block of the target block,
wherein reconstructing the motion vectors comprises:
determining an initial motion vector of a partition of the partitions;
decoding information on a motion vector difference of the partition from a bitstream;
reconstructing the motion vector difference based on the information on the motion vector difference; and
deriving the motion vector of the partition by adding the initial motion vector to motion vector difference,
wherein the information on the motion vector difference includes:
a first syntax element, which is an index for indicating a distance of the motion vector difference among a plurality of pre-defined distances, and
a second syntax element, which is an index for indicating a direction of the motion vector difference among a plurality of pre-defined directions, and
wherein the plurality of pre-defined directions includes two horizontal directions, two vertical directions, and four diagonal directions.
2 . The method of claim 1 , wherein determining the partitioning parameter for the target block comprises:
decoding partition information from bitstream, wherein the partition information includes information related to the initial motion vector indicating a position of a reference region to be referred to for determining the partitioning parameter for the target block;
determining a partitioning structure of the reference region indicated by the initial motion vector, by using pixel values of pixels in the reference region; and
using the partitioning structure of the reference region to determine the partitioning parameter related to a partitioning structure of the target block.
3 . The method of claim 1 , wherein generating of the prediction block of the target block comprises:
for a prediction block corresponding to a first partition split from the target block, determining weights for pixels in the prediction block corresponding to the first partition based on pixel positions that are relative to a boundary between the first partition and a second partition.
4 . The method of claim 3 , wherein determining of the weights for the pixels in the prediction block corresponding to the first partition comprises:
generating a first parameter by using pixels adjacent to the boundary among the pixels in the prediction block corresponding to the first partition;
generating a second parameter by using pixels adjacent to the boundary among the pixels in the prediction block corresponding to the second partition;
determining a boundary type of the boundary among a plurality of boundary types by using a difference between the first parameter and the second parameter; and
determining the weights based on a determined boundary type.
5 . A video encoding method of encoding a target block, the method comprising:
determining a partitioning parameter for the target block;
determining motion vectors which correspond to a plurality of partitions split from the target block by the partitioning parameter;
using the motion vectors to generate prediction blocks corresponding respectively to the partitions;
generating a prediction block of the target block by weighted averaging the prediction blocks corresponding to the partitions;
generating a residual block of the target block based on the prediction block of the target block; and
encoding information on motion vectors corresponding respectively to the partitions, and information on the residual block,
wherein encoding the information on the motion vectors comprises:
determining an initial motion vector of a partition of the partitions; and
encoding information on a motion vector difference of the partition into a bitstream, wherein the motion vector difference of the partition is a difference between a motion vector of the partition and the initial motion vector;
wherein the information on the motion vector difference includes:
a first syntax element, which is an index for indicating a distance of the motion vector difference among a plurality of pre-defined distances; and
a second syntax element, which is an index for indicating a direction of the motion vector difference among a plurality of pre-defined directions, and
wherein the plurality of pre-defined directions includes two horizontal directions, two vertical directions, and four diagonal directions.
6 . The method of claim 5 , wherein determining the partitioning parameter for the target block comprises:
determining a partitioning structure of a reference region indicated by the initial motion vector, by using pixel values of pixels in the reference region; and
using the partitioning structure of the reference region to determine the partitioning parameter related to a partitioning structure of the target block,
wherein information on the initial motion vector is further encoded.
7 . The method of claim 5 , wherein generating of the prediction block of the target block comprises:
for a prediction block corresponding to a first partition split from the target block, determining weights for pixels in the prediction block corresponding to the first partition based on pixel positions that are relative to a boundary between the first partition and a second partition.
8 . The method of claim 7 , wherein determining of the weights for the pixels in the prediction block corresponding to the first partition comprises:
generating a first parameter by using pixels adjacent to the boundary among the pixels in the prediction block corresponding to the first partition;
generating a second parameter by using pixels adjacent to the boundary among the pixels in the prediction block corresponding to the second partition;
determining a boundary type of the boundary among a plurality of boundary types by using a difference between the first parameter and the second parameter; and
determining the weights based on a determined boundary type.
9 . A method for transmitting a bitstream containing encoded data of a target block, the method comprising:
generating the bitstream by encoding the target block; and
transmitting the bitstream to a video decoding apparatus,
wherein generating the bitstream comprises:
determining a partitioning parameter for the target block;
determining motion vectors which correspond to a plurality of partitions split from the target block by the partitioning parameter;
using the motion vectors to generate prediction blocks corresponding respectively to the partitions;
generating a prediction block of the target block by weighted averaging the prediction blocks corresponding to the partitions;
generating a residual block of the target block based on the prediction block of the target block; and
encoding information on motion vectors corresponding respectively to the partitions, and information on the residual block,
wherein encoding the information on the motion vectors comprises:
determining an initial motion vector of a partition of the partitions;
encoding information on a motion vector difference of the partition into a bitstream, wherein the motion vector difference of the partition is a difference between a motion vector of the partition and the initial motion vector;
wherein the information on the motion vector difference includes:
a first syntax element, which is an index for indicating a distance of the motion vector difference among a plurality of pre-defined distances, and
a second syntax element, which is an index for indicating a direction of the motion vector difference among a plurality of pre-defined directions, and
wherein the plurality of pre-defined directions includes two horizontal directions, two vertical directions, and four diagonal directions.