The greyscale image of monsters best demonstrated the worth and success of our work in pattern matching. As the example in the Results section shows, we were able to isolate a single monster as the template and then locate it in the original image as the brightest pixel in the convolved image. Testing our various "convolution" algorithms on this image also revealed important differences in performance. For an increasingly large template size, the FFT proved itself a far superior method in terms of computation time.
Unfortunately, the methods we developed for pattern matching have limited utility. The problem is that a template can be matched (found in an image) only if it already exists in the image, pixel for pixel. In practical situations, this will rarely be the case. One must account for relative size, orientation, and shading, among other things. Thus, even for the specific types of pattern matching which we have studied, there is a great deal of exploration and work to be done.