一种基于立体视邻接帧时空相关性的最小代价函数帧估计算法  

A Smallest Cost Function Frame Estimation Algorithm Based on Spatial-Temporal Correlation in Neighboring Frames of Stereoscopic Sequences

作　　者：骆艳[1] 张兆扬[1] 

机构地区：[1]上海大学通信学院电子信息系,上海200072

出　　处：《电子学报》2003年第10期1513-1517,共5页Acta Electronica Sinica

基　　金：国家自然科学基金 (No .69972 0 2 7)

摘　　要：为了在立体视频序列编码中获得高的压缩率 ,需要对立体视频序列中一个视的序列按传统方法进行独立编码 ;另一个视的序列中 ,只对其中一些参考帧 (I帧或P帧 )按视差补偿预测的方法进行编码 ,其余帧不进行编码和传输 ,而在解码端用立体视帧估计的方法得到重建 .本文提出了一种基于立体视中邻接帧在图像、视差场和运动矢量场之间高度相关性的方法 .对于因遮挡而缺乏估计的区域 ,则结合了图像强度的连续性和运动 ,视差矢量的分布特性 ,构造了代价方程并估计出该部分的运动矢量及强度值 .实验证明 ,重建出来的帧图像在视觉和信噪比意义上均具有较好的效果 .In order to achieve high compression ratio in coding stereo video sequences,one sequence of stereoscopic sequences is independently coded by a conventional MPEG type scheme,only a few reference frames in another sequence are coded.The rest of frames,however,are neither coded nor transmitted.They are reconstructed at the decoder using a stereoscopic frame estimation technique.This paper studies the highly spatial and temporal correlation between images,disparity fields and motion vector fields in the neighbor frames of stereoscopic sequences and proposes a frame estimation technique to explore these correlations.As to those un estimated regions due to occlusion,this paper also constructs a cost function combined with the continuity of image intensity and the distribution characteristic of motion and disparity vector to estimate the motion vectors and intensities of them.The experiment results show that this frame estimation algorithm has good results on view perception and PSNR.

关 键 词：立体视频 遮挡 帧估计 

分 类 号：TN911.1[电子电信—通信与信息系统]