Using feature probabilities to reduce the expected computational cost of template matching

Abstract Matching of two digital images is computationally expensive, because it requires a pixel-by-pixel comparison of the pixels in the image and in the template for every location in the image. In this paper we present a technique to reduce the computational cost of template matching by using probabilistic knowledge about local features that appear in the image and the template. Using this technique the most probable locations for successful matching can be found. In the paper we discuss how the size of the features affects the computational cost and the robustness of the technique. We also present results of experiments showing that even simple methods of feature extraction and representation can reduce the computational cost bymmore than an order of magnitude.