Speed-up Template Matching through Integral Image based Weak Classifiers

Template matching is a widely used pattern recognition method, especially in industrial inspection. However, the computational costs of traditional template matching increase dramatically with both template-and scene imagesize. This makes traditional template matching less useful for many (e.g. real-time) applications. In this paper, we present a method to speed-up template matching. First, candidate match locations are determined using a cascaded blockwise computation of integral image based binary test patterns. Then, traditional template matching is applied at the candidate match locations to de-termine the best overall match. The results show that the proposed method is fast and robust