Granted Patent B2

US 12,425,620 · App. 18/238,340 · Granted Sep 23, 2025

Inventors: Yinji Piao (Suwon-si, KR); Kyungah Kim (Suwon-si, KR); Minsoo Park (Suwon-si, KR); Minwoo Park (Suwon-si, KR); Kwangpyo Choi (Suwon-si, KR)

Assignee: SAMSUNG ELECTRONICS CO., LTD.

H04N19/44H04N19/105H04N19/137H04N19/176

Abstract

Claims (29)

1. An image decoding method comprising:

identifying at least two neighboring sub-blocks that have a same motion vector and are adjacent to each other from among a plurality of neighboring sub-blocks that are adjacent to a current block;

determining a representative motion vector and representative location information corresponding to the at least two neighboring sub-blocks;

obtaining an initial parameter of a model for determining a motion vector from a block reconstructed in a sub-block unit mode prior to the current block;

refining a parameter of the model based on the initial parameter, the representative motion vector and the representative location information;

determining a motion vector of a current sub-block included in the current block based on location information of the current sub-block and the parameter; and

predicting the current block based on the motion vector of the current sub-block,

wherein the representative motion vector is determined based on the same motion vector, and

wherein the representative location information indicates coordinates that are determined based on the at least two neighboring sub-blocks.

2. The image decoding method of claim 1 , wherein the coordinates correspond to at least one from among a center of the at least two neighboring sub-blocks or a center of a neighboring block comprising the at least two neighboring sub-blocks.

3. An image decoding method comprising:

identifying at least two neighboring sub-blocks that have a same motion vector and are adjacent to each other from among a plurality of neighboring sub-blocks that are adjacent to a current block;

determining a representative motion vector and representative location information corresponding to the at least two neighboring sub-blocks;

determining individual motion vectors and pieces of individual location information corresponding to neighboring sub-blocks that have different motion vectors or are not adjacent to each other from among the plurality of neighboring sub-blocks;

obtaining an initial parameter of a model for determining a motion vector from a block reconstructed in a sub-block unit mode prior to the current block;

refining a parameter of the model based on the initial parameter, the representative motion vector, the representative location information, the individual motion vectors, and the pieces of individual location information;

determining a motion vector of a current sub-block included in the current block based on location information of the current sub-block and the parameter; and

predicting the current block based on the motion vector of the current sub-block.

4. An image decoding method comprising:

comparing motion vectors of a first neighboring block and a second neighboring block that are adjacent to each other from among neighboring blocks that are adjacent to a current block;

obtaining an initial parameter of a model for determining a motion vector from a block that is reconstructed in a sub-block mode prior to the current block;

based on a result of the comparing indicating that a first motion vector of the first neighboring block and a second motion vector of the second neighboring block are a same motion vector, determining representative location information corresponding to the first neighboring block and the second neighboring block, and refining a parameter of the model based on the initial parameter, the representative location information and one of the first motion vector and the second motion vector;

based on the result of the comparing indicating that the first motion vector and the second motion vector are different from each other, refining the parameter based on the initial parameter, the first motion vector, the second motion vector, first location information about the first neighboring block, and second location information about the second neighboring block;

determining a motion vector of a current sub-block included in the current block, based on location information about the current sub-block and the parameter; and

predicting the current block based on the motion vector of the current sub-block.

5. The image decoding method of claim 4 , wherein the representative location information indicates coordinates corresponding to a center of the first neighboring block and the second neighboring block.

6. The image decoding method of claim 4 , wherein the first location information indicates coordinates corresponding to a center of the first neighboring block, and

wherein the second location information indicates coordinates corresponding to a center of the second neighboring block.

7. The image decoding method of claim 4 , wherein at least one of the first neighboring block or the second neighboring block is a coding unit or a sub-block unit included in the coding unit.

Assignments (1)

Priority Claims (2)

Continuity (2)

References Cited (31)