An Estimation Optimization Algorithm of Side Information Based on General Prediction

摘要

The coding efficiency of the Wyner-Ziv video decoder is largely dependent on the quality of the side information obtained in the decoder. The reason why constructing effective side information is difficult is partly caused by unavailability of the original video sequence in the decoder. The traditional motion search method is used to obtain side information, which greatly increases the complexity of Wyner-Ziv video decoding. This paper proposes a new method of extracting side information based on the general prediction idea. This new method is called Wyner-Ziv Video Coding (WZUP) with general prediction. This method does not conduct motion search in the decoder or presuppose a model for the original input video sequence. Instead, side information is estimated based on observations of video data that have been reconstructed in the past. Through testing, this method can greatly reduce the decoder's decoding complexity and achieve good side information estimation performance. This makes it possible to design video encoders and decoders with low computation complexity.