On reducing the size of structured meshes with a novel video object extraction algorithm

Mesh-based motion analysis is a popular inter-frame video compression model. Compression with this model is further increased when the size of the mesh used to represent video frames is reduced. In this paper, a video object extraction algorithm that reduces the size of structured meshes is proposed. It uses block-matching analysis to extract video objects (or dynamic contents) in video frames, and subsequently reduces the mesh size by removing the nodes that represent the video background (or motionless contents). In essence, only video objects are represented in the reduced mesh. During frame reconstruction, video objects are recovered from the affine transform, while the video background is mapped directly from the reference frame. Our study shows that the proposed algorithm can reduce on average 77.9% of the mesh nodes per frame for the Claire sequence (which has a motionless background). Also, the frame reconstruction quality is not affected as a result of using the proposed algorithm. Therefore, we claim that the proposed algorithm is effective in achieving high video compression.