Research on effective coding algorithm for multi-view video

摘要

Based on deeply analyzing and studying of multi-view video coding technology in JMVC (Joint Multi-view Video Coding), an effective coding algorithm in views of model decision and fast block searching was put forward in this paper. First, we use different search algorithms for interview prediction and frame prediction, respectively, to reduce search time and improve search efficiency. Second, we analyze the relationship among the model of current macroblock, the model of adjacent coded macroblocks and the corresponding position of reference frame. According to the result we can optimize the model decision algorithm for current coding macroblock, and reduce the search time for refined small block. The improved fast search algorithm is effective to the sequences containing fast motion area, and the improved model decision algorithm is effective to the sequences containing slow motion. Therefore, we combine the two improved algorithms, and apply them to sequences with different characteristics. The experimental results show that the improved algorithms get significant improvement in coding effect. Compared with the original algorithm in JMVC, the improved algorithm can reduce the coding time by 84.5% under the premise of insuring reconstruction video quality and coding overhead. (C) 2015 Elsevier GmbH. All rights reserved.