基于预搜索的高效双目分形视频编码

摘要

An efficient stereo fractal video coding based on pre-searching was presented for the stereo video coding successfully. The basic stereo fractal video coding was improved by using tree structure macroblock partition, Discrete Cosine Transform(DCT) based encoding for original frame, reducing block searching range and computational repetition in advance. In stereo fractal video coding, the left view video was arranged as a main view and the Motion Compensation Prediction (MCP) was taken as coding structures by employing pre-searching limitation condition, improved UMHexagonS motion estimation algorithm, deblocking loop filter and sub-pixel block matching. The right view video was arranged as an enhanced view, the MCP and Disparity Compensation Prediction(DCP)were combined to be as coding structures, and the matching block with smallest matching error was chosen as the estimation results. In the DCP coding, a fast disparity estimation algorithm was proposed by making full use of disparity distribution restriction conditions. Experimental results indicate that the compression time by proposed stereo fractal video coding algorithm is 18%~23% that by the basic stereo fractal video coding algorithm, and the compression ratio has improved by 15. 13-47. 49 while a proper Peak Signal Noise Ratio(PSNR) is maintained . These results show the compression performance of the stereo fractal video coding algorithm and its applications are more flexible and practical.%设计了基于预搜索的高效双目分形视频编码,并成功应用于立体视频编码之中.对基本分形双目视频编码进行了改进.利用了树状划分准则,起始帧采用块离散余弦变换(DCT)编码,简化了块搜索范围并提前减少了重复运算.在双目立体视频编码中,以左通道为基本层,采用单独的运动补偿预测( MCP)方式进行编码,充分利用了预搜索限制条件、改进的非对称十字形多层次六边形格点搜索算法、去方块环路滤波和分数像素块匹配算法；以右通道为增强层,采用MCP加视差补偿预测(DCP)方式进行编码,选择误差最小的匹配块作为预测结果.在进行DCP编码时,充分利用视差分布约束条件,提出了快速的视差估计算法.实验结果表明,提出的编码方法在保证一定的峰值信噪比(PSNR)前提下,平均压缩时间是基本分形双目视频编码的18％～23％,压缩比提高了15.13～47.49,显著地改善了基本分形视频压缩算法的性能,使分形视频压缩的应用具有更大的灵活性和实用性.