Segmentation of visual images: experimental data and modeling

Abstract

Experimental data on the segmentation of simple geometrical figures surrounded by a frame are compared for the first time to our knowledge with the results of a model of modules that filters the images in local areas of the visual field. It is shown in the experiments that images are perceived as two separate images when they reach a definite spacing between them, depending on their size and shape. When they are small, the distances are comparable with the optical point-spread function and the size of the highest-frequency receptive fields of the neurons in the primary visual cortex, and, when they are large, with the size of modules that optimally describe the images (with the maximum energy conservation in the images when there is a limited number of filters). The images are segmented when the second image (the frame) goes beyond the limits of the module. The results were confirmed by our earlier data on the study of the Oppel–Kundt illusion and by estimating the width of the spatial intervals.