基于小波变换的多分辨匹配算法:

首先利用小波的多分辩率特性将匹配图像和模板图像分解到乙层上，并且只保留LL低频部分，然后利用归一化相关法作为相似性度量，进行由粗到细的相关匹配过程，每次利用下一层的匹配结果在上层小范围内搜索。这样做极大地减少了搜索空间，而且减少了每次匹配计算相似度的数据量。

-Wavelet multiresolution matching algorithms: a wavelet multi-resolution feature will match the image and the template image is decomposed to the B layer, and retain only low-frequency part of the LL, and then use normalized correlation as a similarity measure, carried out by the rough to the relevant matching process, each time using the matching results of the next layer in the upper small scale search. This greatly reduces the search space, but also reduced the calculation of similarity for each match the amount of data.