Selection of image primitives for general-purpose visual processing

摘要: The results of psychophysical experiments suggest the existence a hierarchy visual features based on relations between image contours. human system appears to be preattentively, selectively sensitive contours which contain certain these features. These can used develop computer vision algorithms for: (1) selective enhancement correlate with their perceptual significance; and (2) segmentation boundary images into sets have high probability depicting single object. extreme simplicity algorithms, as well ability generate perceptually significant results, demonstrate advantages using uncover invariants structure physical world that is value for perception.