Digital data storage is becoming ever more abundant and cheap. This, along with other technological advances, has brought about an age of mass storage of information, much of it in the form of images. In order to be able to process these stockpiles of image data, new and faster computer algorithms are needed.
One area of interest is that of image mosaicking, i.e., comparing two overlapping images and finding the proper scaling, angle of rotation, and translation needed to fit one with the other. Early methods for mosaicking images included visual inspection or exhaustive, pixel by pixel, search for the best match. With such large quantities of images as we have today, and the increasing need for accuracy, these methods are no longer feasible. Several new mosaicking methods have been proposed based on Fast Fourier Transform (FFT).
The existing FFT-based algorithms do not always lead to reasonable mosaicking. In this paper, we formalize the corresponding expert rules and, as a result, design an optimal FFT-based mosaicking algorithm.