An empirical study of selected approaches to the detection of edges in noisy digitized images

This paper is concerned with an empirical evaluation of the relative performance of several selected edge detectors for the detection of edge structure in noisy digitized images. In particular, we consider both the Hueckel operator and a recently introduced class of edge detectors implemented as two-dimensional infinite inpulse response (IIR) or recursive digital filters. The latter were originally developed on the basis of least mean-square Wiener spatial filtering concepts for an assumed stochastic model of edge structure in typical imagery data. These two schemes are compared with respect to performance, computational complexity and suitability for specific applications. Problems in making these comparisons are discussed. The implementations of both operators used for this comparison are described in detail.