基于小波变换的图像匹配算法

Stereo Matching Algorithm Based on WaveletTransform

为了寻找立体像对中的对应点并求出视差，采用Ｍａｌｌａｔ的多尺度小波分析算法，在对图像进行多级分解的基础上进行逐级匹配，并且扩展了传统的视差定义，将两个限制条件：视差连续性和方向一致性融合进视差的定义当中，因而在视差求取的同时也完成了粗差剔除的过程，提高了算法的效率．实验证明，此方法在计算时间和精度上都是切实可行的．

The corresponding points in a pair of im ages w ere found by decom posing im ages into several layers based on m ultiscale analysis(wavelet transform ). Also, the definition of disparity w as extended w ith inclusion ofthe tw o constraints: continuity ofdisparity and consistency oforientation. So disparity and error exclusion can be finished sim ultaneously and the efficiency ofthis algorithm is im proved. The experim ents show thatthis algorithm is feasible both in com puting tim e and precision.