A template extraction approach for image recognition

A robust approach for extracting template from images with relatively poor quality is presented in this paper. The approach combines fast Fourier transformation algorithm and weighted normalized cross correlation algorithm to deal with images taken under weak template. Basically, it consists of three steps: 1) searching candidate template from the target image using prefabricated template, and 2) determining the count of template matching area and adjusting the candidate template by using neighborhood growth method, and 3) evaluating the optimal template among the candidate templates. A set of experiments has been performed and the results shows that the proposed approach can obtain templates accurately and robustly.