Fundamental Bounds on Edge Detection : An Information

We treat the problem of edge detection as one of statistical inference. Local edge cues, implemented by lters, provide information about the likely positions of edges which can be used as input to higher-level models. Diierent edge cues can be evaluated by the statistical eeectiveness of their corresponding lters evaluated on a dataset of 100 pre-segmented images. We use information theoretic measures to determine the eeective-ness of a variety of diierent edge detectors working at multiple scales on black and white and colour images. Our results give quantative measures for the advantages of multi-level processing, for the use of chro-maticity in addition to greyscale, and for the relative eeectiveness of diierent detectors.