采用小波变换的立体匹配:一种基于相位的方法

Stereo Matching Using Wavelets Transform: A Phase-based Approach

本义提出了一种采用小波变换基于相位的立体匹配算法。该算法引入多分辨率分析思想，借助小波变换的Ｍａｌｌａｔ算法，采用金字塔式的多尺度匹配结构，从而显著地提高了匹配效率。同时，该算法利用平滑性约束条件实现了相位不稳定区域视差的自适应插值，保证了良好的匹配效果。立体象对的测试结果表明了该算法的有效性。

A phase-based stereo matching algorithm using wavelets transform is presented in this paper The idea of multiresolution analysis is introduced and a multi-scale pyramid matching structure from Mallat's algorithm of wavelets transform is adopted, so as to improve the matching efficiency. At the same time, the adaptive interpolation of disparities is implemented in some areas where phase is unstable by introducing the smooth constrain, and good matching effects are ensured. The validity of this algorithm is shown by the results from stereo image pairs.