This thesis presents a new, simple and effective low level processing method to obtain meaningful features such as edges and corners. In both edge and corner extraction algorithms, we use two oriented cross operators called COP(Crosses as Oriented Pair).
First, we propose a new edge detection algorithm. To obtain meaningful edge, many conventional edge operators use derivative convolution masks and are followed by the conventional non-maximum suppression algorithm which needs rather complicated edge direction calculation. Moreover, most conventional derivative-based operators suffer from poor connectivity at junctions, sensitivity to noise and two extrema when they are applied to a line. But COP makes it possible to find edge direction very easily, localize edge position accurately and obtain connected edges.
Second, we propose a new corner detection algorithm using COP. Most conventional corner detectors have shortcomings such as missing junctions, poor localization, sensitivity to noise and high computational cost. With the useful characteristics of the COP and simple rules, we can accomplish a very fast, accurate and robust corner detection than any other corner detector. Performances of two proposed algorithms are described with test results.