Improved method of rotation invariant template matching using ROC

Abstract

The orientation code is a quantization of pixel gradient orientation. It has the most important characteristic of illumination invariant and has become an important template matching method. In order to improve accuracy and real time ability of the original algorithm, we present a new conception of relative orientation code (ROC), and design template matching method based on the ROC. This method first selects candidate target points using histogram matching, then calculate the rotation angle of candidate sub-image and rotate the sub-image, and finally match the in the sub-image. The experiment shows that this method is robust when the image is rotated by any angle. The candidate points are selected and matched accurately. The computing time is less than the original method.