A fast template matching algorithm based on central moments of images

This paper presents a new template matching algorithm in which the central moments are used as the matching features and the idea of integral image are used to decrease the time for image matching process. The algorithm also combines coarse-fine two-stage searching methods to effectively solve the problem of finding the peak point of the correlation functions accurately and robustly. It is different with traditional methods in which pixels are used as matching features. Experiment results show that the image process procedure can be finished within 40 ms, which is the rate of image signal send from TV camera for real time object tracking. The results has broad applications in the fields of real time moving object tracking, pattern recognition and the machine vision, etc.