Object Searching with Combination of Template Matching

Abstract

Object searching is the identification of an object in an image or video. There are several approaches to object detection, including template matching in computer vision. Template matching uses a small image, or template, to find matching regions in a larger image. In this paper, we propose a robust object searching method based on adaptive combination template matching. We apply a partition search to resize the target image properly. During this process, we can make efficiently match each template into the sub-images based on normalized sum of squared differences or zero-mean normalized cross correlation depends on the class of the object location such as corresponding, neighbor, or previous location. Finally, the template image is updated appropriately by an adaptive template algorithm. Experiment results show that the proposed method outperforms in object searching.