FIELD: physics.
SUBSTANCE: invention relates to video encoding and decoding. Methods, devices, systems, architectures and interfaces are disclosed for improving prediction with motion compensation at the sub-block (SB) level based on motion vector (MV) refinement. Decoding method includes receiving a bit stream of encoded video data, the bit stream including at least one block of video data including a plurality of SBs; MV derivation, including a decoder-based MV process (DMVR), for at least one SB in a unit for generating a refined MV for each SB; SB-based motion compensation in at least one sub-block to generate SB-based prediction within each SB; obtaining a spatial gradient for prediction within each SB; determining the MV offset of each pixel in each SB; obtaining parameters of variation of intensity in each SB based on spatial gradients and displacements of MV using an optical flow equation; and updating the prediction within each SB based on the obtained intensity variation parameters.
EFFECT: high coding efficiency.
21 cl, 13 dwg
Authors
Dates
2024-05-28—Published
2020-03-26—Filed