Hybrid object-based/block-based coding in video compression at very low bit-rate

作者:

Highlights:

摘要

In this paper, an algorithmic architecture of video coding at very low bit-rate is presented. This architecture incorporates block- and object-based coding algorithms into the same coder and manages them in a cooperative way: I and P frames are coded by block-based algorithm and B frames are coded by bi-directional object-based algorithm. An extra degree of freedom in video coding is introduced to achieve higher compression ratios while maintaining good visual quality. We give details of the object-based coding algorithm which is based on a source model of rigid 2-D objects with 3-D motions. Each moving object in B frame is coded with its contour and a pair of motion parameter sets (forward and backward motions). The bi-directional object-based motion parameters are obtained by a hierarchical estimation. The object image is interpolated using two neighboring pictures, which are block-based coded. ITU-T H.263 is chosen as the block-based coding scheme and the simulations are performed on MPEG-4 test sequences, which contain low-to-medium spatial detail and low amount of movement, but do not restrict to head_and_shoulder images. Experiment results show that the coder provides higher image quality with smooth and satisfactory motion.

论文关键词:Object-based image coding,Bi-directional object-based image coding,Motion segmentation,Hierarchical block matching

论文评审过程:Available online 19 May 1998.

论文官网地址:https://doi.org/10.1016/S0923-5965(97)00023-4