QUALCOMM Incorporated
Adaptive affine motion vector coding

Abstract:

A video coding device is configured to determine control point motion vectors (CPMVs) for an affine motion predicted block of video data. For a four parameter affine model, two CPMVs may be positioned at any two corners of the current block, such as top-left and bottom-right or bottom-left; top-right and bottom-left or bottom-right; or bottom-left and bottom-right. For a six parameter affine model, three CPMVs may be positioned at any three corners of the current block, such as top-left, bottom-left, and bottom-right; top-right, bottom-left, and bottom-right; or top-left, top-right, and bottom-right. The video coder may further predict the CPMVs using motion information of neighboring blocks to the determined CPMVs, code the CPMVs, and code the current block using the CPMVs. The video coder may be a video encoder or a video decoder.