Adaptive overlapped block prediction in variable block size video coding

A method, comprising:

• encoding a current block in a current frame of a video stream, comprising:
• generating overlapped predictions for each previously encoded adjacent block using at least two discrete overlap region sizes, including:
• determining a first overlap region within the current block and adjacent to a previously encoded first adjacent block, wherein at least one side of the current block is adjacent to two or more previously encoded adjacent blocks;
• and
• generating an overlapped prediction of pixel values in the first overlap region as a weighted function of:
• a first prediction using first prediction parameters of the current frame, and
• a second prediction using second prediction parameters of the first adjacent block;
• determining a rate distortion cost for each of the overlapped predictions;
• selecting an overlap region size for all overlapped predictions based on rate distortion cost optimization;
• and
• encoding a weighted function index that indicates the selected overlap region size,
• wherein the weighted function is adaptable by a function of a difference between the first prediction parameters and the second prediction parameters, and the weighted function is defined such that, when the first prediction parameters and the first adjacent block are unchanged, the weighted function results in a first overlapped prediction for the first overlap region when the second prediction parameters have first values and the weighted function results in a second overlapped prediction for the first overlap region when the second prediction parameters have second values different from the first values.

A method of encoding a current block of a current frame in a plurality of frames of a video stream, comprising:

• generating a first prediction block of pixel values in the current block based on a first prediction using first prediction parameters;
• defining an overlap region within the current block and adjacent to an adjacent block among a plurality of previously encoded blocks having variable block sizes, wherein at least one side of the current block is adjacent to two or more of the previously encoded blocks;
• generating an updated prediction of pixel values for pixels within the overlap region, the updated prediction being a weighted function of the first prediction and a second prediction, the second prediction being based on second prediction parameters, wherein the weighted function is adaptable by a function of a difference between the first prediction parameters and the second prediction parameters such that when the first prediction parameters and the second prediction parameters indicate that one of the current block or the adjacent block is intra-predicted, the weighted function defines a size of the overlap region according to a predetermined function of a size of the current block, and such that when the first prediction parameters and the second prediction parameters indicate that each of the current block and the adjacent block is inter-predicted, the weighted function defines the size of the overlap region based on comparing the first prediction parameters and the second prediction parameters;
• and
• encoding an index indicating a type of the weighted function used for the overlapped prediction, wherein the index indicates one of a raised cosine window function or a weighted function based on relative distance to block center points of the current block and the adjacent block.