ABSTRACT
This chapter discusses the problem of image matching or locating a known part of the image, which is usually described as a pattern. This type of problem is typical in applications, such as the search for reference points in stereo vision, the location of a certain object in a scene, or the tracking of objectives in a sequence of images. The fundamental idea of image comparison (Template Matching) is simple. The pattern (template) to be found in the image is moved over all the pixels of the image as if it were a linear filter. Then, a similarity evaluation between the pattern and the pixels is measured. Obviously, the determination of similarity between the pattern and each of the points of the image is not easy since the pattern could appear in the image scaled, rotated, or distorted.
