Robust Video Coding Techniques for Wireless Video Communications

摘要

The objective of this SBIR project is to develop a video coding technique that isrobust for wireless video communications in terms of video quality degradation as channel noise increases. Phase I of the project is a feasibility study of several innovative error-handling techniques based on the framework of vector transform coding, which has been shown to provide much better compression performance than the standard techniques for still image and video compression. As a result of the phase I project, three different index assignment techniques for vector quantization in vector wavelet domain have been studied. The three techniques are (1) optimal index assignment without knowing the code-vector distribution; (2) optimal index assignment with known code-vector distribution; and (3) joint optimization of index assignment and codebook design. Compared with the standard video coding method with and without error correction coding, all three new techniques have shown much more robust performance in a wide range of channel noise levels. Among the three new techniques, we have concluded that the second one is the best, which provides better performance than the first one and about the same performance as the third one but with lower complexity.