Method for coding and decoding a block of pixels from a motion model

A device for coding a block of pixels of an image comprising at least a processor configured to:

• determine a prediction block of a current block from at least one motion model defined by a set of parameters, said motion model associating a motion vector pointing at a pixel of a prediction block with each pixel of the current block,
• calculate a residual block between said current block and said prediction block, said motion model associating a motion vector pointing at a pixel of the prediction block with each pixel of the current block, and code said residual block;
• wherein said processor is further configured to:
• segment the causal neighborhood of the current block into a plurality of sets of blocks, a set of parameters being determined for each one of the sets of blocks motion data associated with the plurality of sets of blocks of the causal neighborhood comprise motion vectors, wherein the set of parameters of the motion model are determined by minimization to a least squares of the sum of differences between each one of said motion vectors of the plurality of sets of blocks of the causal neighborhood and the associated motion vector resulting from the motion model;
• determine a prediction block of the current block with each one of said sets of parameters;
• calculate a residual block between the current block and each one prediction block and code said residual blocks;
• and
• select, among coded residual blocks, the coded residual block to transmit according to a predefined selection criterion and generate information identifying the set of parameters having served to determine the prediction block associated with the selected residual block, said information being transmitted with said coded residual block;
• wherein a first minimization to the least squares is applied on the a set of blocks then a second minimization to the least squares is applied on a second set of blocks then a third minimization is applied on the whole causal neighborhood to determine the parameters of the motion model.

A method for decoding a block of pixels of an image from an associated coded residual block comprising:

• decoding said coded residual block;
• determining a prediction block of a current block from a motion model defined by a set of parameters, said motion model associating a motion vector pointing at a pixel of a prediction block with each pixel of the current block;
• determining the current block from said decoded residual block and said prediction block, wherein, for said current block, said prediction block is determined by:
• segmenting a causal neighborhood of the current block into a plurality of sets of blocks, a set of parameters being determined for each one of the plurality of sets of blocks, motion data associated with the plurality of sets of blocks of the causal neighborhood comprise motion vectors, wherein the set of parameters of the motion model are determined by minimization to a least squares of the sum of differences between each one of said motion vectors of the plurality of sets of blocks of the causal neighborhood and the associated motion vector resulting from the motion model;
• decoding information identifying a set of parameters;
• determining a prediction block of the current block from a motion model defined by said set of parameters identified by said decoded information;
• wherein a first minimization to the least squares is applied on the first set of blocks then a second minimization to the least squares is applied on the second set of blocks then a third minimization is applied on the whole causal neighborhood to determine the parameters of the motion model.

A device for decoding a block of pixels of an image from an associated coded residual block comprising at least a processor configured to:

• decode the residual block;
• determine a prediction block of a current block from a motion model defined by a set of parameters;
• and
• determine the current block from the decoded residual block and said prediction block, said motion model associating a motion vector pointing at a pixel of the prediction block with each pixel of the current block, wherein the processor is further configured to:
• segment the causal neighborhood of the current block into a plurality of sets of blocks, a set of parameters being determined for each one of the sets of blocks, motion data associated with the plurality of sets of blocks of the causal neighborhood comprise motion vectors;
• decode information identifying a set of parameters, wherein the set of parameters of the motion model are determined by minimization to a least squares of the sum of differences between each one of said motion vectors of the plurality of sets of blocks of the causal neighborhood and the associated motion vector resulting from the motion model;
• and
• determine a prediction block of the current block from a motion model defined by said set of parameters identified by said decoded information;
• wherein a first minimization to the least squares is applied on the first set of blocks then a second minimization to the least squares is applied on the second set of blocks then a third minimization is applied on the whole causal neighborhood to determine the parameters of the motion model.