Image segmentation at contour intersections in human focal vision.

It would be efficient for the visual image to be broken initially into small segments, each of which could be analyzed separately before being related to the others to build a representation of the whole. The places in the image where such segmentation should occur will depend on the image and will not be fixed points in retinal space. In this study, it is shown that high-precision-shape information cannot be integrated across contour intersections. It is suggested that contour intersections are a particular feature of the retinal image at which segmentation inevitably occurs.