ABSTRACT
One of the major tasks of the human visual system is to extract quickly and effortlessly the most salient information from an image to form a symbolic representation of the scene. This chapter outlines the details of the model and illustrates its application to several interesting images. It shows how the model successfully predicts human perception, both qualitatively and quantitatively, in conditions where most previous models fail. Many models have been proposed for line and edge detection, all similar in one major respect: they convolve the image with simple linear operators of various sizes and search for maxima or zero-crossings in the output. However, this approach runs into several difficulties, particularly when the image features are adjacent and when the features are not simply edges or lines but combinations of both.
