快速分形立体视频编码系统的设计

提出了一种基于分形视频编码的快速立体视频编码算法。首先,对传统分形视频编码方法进行了改进:采用基于DCT变换的方式对I帧图像进行编码,同时采用树状划分方法对非I帧图像进行块匹配。在立体视频编码中以左通道为基本层,右通道为增强层;左通道采用单独的运动补偿预测方式(MCP)进行编码,右通道采用MCP加视差补偿预测方式(DCP)进行编码。在进行DCP编码方式时,充分利用立体平行摄像结构中的偏振性和方向性简化DCP搜索方式,由此提出了一种快速搜索算法。实验结果表明,在保证峰值信噪比(PSNR)和压缩比(CR)基本不变的前提下,本文所提出的快速编码算法能够将运算复杂度降低为全搜索算法的0.028～0.029倍,增强了立体视频编码的实用性。

Design of fast fractal stereo video coding system

A fast stereo video coding algorithm based on fractal video coding was presented in this paper. Firstly, the conventional Circular Prediction Mapping/Non-contractive Interframe Mapping(CPM/NCIM) method was improved,in which frame I of the video sequences was coded based on Discrete Cosine Transform(DCT) and the others were partitioned into subblocks by a tree structured partition method . In the stereo video encoding, the left channel was a base layer and the right channel was a enhanced layer. The left channel was encoded by using the Motion Compensation Prediction(MCP) method,and the right channel was encoded by using both MCP and Disparity Compensation Prediction(DCP) methods. In the encoding of DCP, the polarization property and the epipolar line property in the parallel stereo encoding structure was used of to simplify the DCP search method. Therefore, a fast DCP encoding method was presented. The experiment results indicate that the computational complexity by the proposed method has reduced to 0.028-0.029 times comparing to that of the full search method while the reconstructed image quality is as good as that in the full search,which makes the stereo video encoding method more useful.