Multiview Video Coding Algorithm Analysis And Optimization

Multi-view video are a group of video signals that simultaneously captured by a group of cameras from different angles of the same scene,which includes a large number of 3D scene information. Presenting to people in a strong three-dimensional world, while providing interoperability features, users can swap the views to watch contents of the different views according to their preferences at any time, this initiative brings people an unprecedented experience. However, because of vast amounts of data, multi-view video has a high demand to storage capacity and channel bandwidth, and the demand hinder its development. In recent years,How to improve the efficiency of multi-view video coding has become the focus of the research, On the premise of improving the coding efficiency, in order to meet the growing need for real-time webcast, we need to reduce algorithm computational complexity and save the encoding time.This article is based on the starting point, after the analysis of the principle of fast search algorithm TZSearch in multi-view video, Based on this, proposing an optimal algorithm for multi-view video coding, which has mainly optimized on three aspects: first, selecting the optimal search model to search the optimal matching points; secondly, flexible search strategy: employing raster search to video sequences that have strong correlation between views; employing a series of hexagonal with different steps to search video sequences that have strong correlation in the time domain; finally, in order to quit the search in time after matching the optimal point, some thresholds are set.Experimental results show that the optimized algorithm effect is obvious: on the premise of guaranteeing the quality of reconstruction video and coding overhead remains unchanged, greatly reducing the encoding time. According to various characteristics of sequences for the test (multi-view video sequences captured by parallel cameras), the encoding time of algorithm reduced to a quarter of the original, and partial sequences have better effect, can reduce 80% of the encoding time.