多视点视差估计中基于图像块特征矢量和相似尺度的参考视点选择方法

Reference View Selection Based on Characteristics of Image Block and Similarity Measurement in Multi-view Disparity Estimation

在多视点视频处理中,采用多参考视点的视差估计方法能够在很大程度上消除各个视点间的冗余性,达到对多视点视频数据有效编码压缩的目的。研究了多视点信号处理中视差估计的多参考视点选择问题,采用快速Walsh变换谱构造图像块特征矢量,并提出了基于图像块特征和相似尺度的多参考视点选择方法。实验结果证明,本文方法能在保证编码质量的前提下,减少视差估计的计算复杂度。

Disparity estimation is used to eliminate redundancy among different views in multi-view video coding in order to obtain high compression efficiency. In this paper, characteristics of image block are extracted by using fast Walsh transformation, and a new method for reference view selection in multi-view disparity estimation is presented, based on the characteristic vector of image block and its similarity measurement. Experimental resuhs show that the proposed method can reduce the computational complexity in multi-view video coding with almost the same quality of reconstructed images.