Silhouette area based similarity measure for template matching in constant time

Abstract

In this paper, we present a novel, fast, resolution-independent silhouette area-based matching approach. We approximate the silhouette area by a small set of axis-aligned rectangles. This yields a very memory efficient representation of templates. In addition, utilizing the integral image, we can thus compare a silhouette with an input image at an arbitrary position in constant time. Furthermore, we present a new method to build a template hierarchy optimized for our rectangular representation of template silhouettes. With the template hierarchy, the complexity of our matching method for n templates is O(logn). For example, we can match a hierarchy consisting of 1000 templates in 1.5 ms. Overall, our contribution constitutes an important piece in the initialization stage of any tracker of (articulated) objects. © Springer-Verlag Berlin Heidelberg 2010.