Object based video coding
Md. Shamim · Library and Archives Canada (Government of Canada) · 2000
In this thesis, I propose a method to segment moving objects in image sequences, so that each can be represented by its boundary and a compact motion description. I also propose a new binary tree coding technique to code motion boundaries in an efficient way. Assuming that a scene consists of a small number of moving objects, each pair of frames is processed as follows: First a small number of "movement classes" is recursively identified, each represented by two or more motion parameters. Second a spatially segmented version of the reference frame (i.e. the later of the two frames) is used to classify segments into movement classes. Third segments using various similarity heuristics are merged together. Finally the motion boundaries and vectors are efficiently coded. Experimental results of some standard test sequences have shown that the proposed algorithm results in good quality motion segmentation with a small number of motion vectors. The coding of motion boundaries uses a modified binary tree coding algorithm, which shows better performance than conventional quadtree, binary tree and READ coding algorithms applied on practical motion boundary images.