Subject of study. The fundamental mechanisms of visual image processing, such as size estimation and segmentation, are considered. The possibility of describing these mechanisms using the model of modules proposed by V. D. Glezer and based on electrophysiological data from studies of the receptive fields of neurons in the visual cortex was investigated. Aim of study. This study aimed to investigate mechanisms of size estimation and the segmentation processes by comparing the experimental data obtained by the authors with modeling results. Methods. Psychophysical methods were used for the experimental part. Two different paradigms were used for size estimation: a modified Ebbinghaus illusion with different surroundings of the test stimuli or a comparison of the sizes of upright and oblique crosses previously used in experiments on segmentation. Spatial-frequency filtering of images in local areas of the visual field was performed during modeling using a finite set of filters. Main results. The dependence of the size estimate on the distance between images and on their shape and the relationship of those estimates with segmentation was demonstrated for the first time to our knowledge. Another new result shows that images with different shapes are perceived as equal in size if the modules that optimally describe these images (i.e., maximum energy is preserved in images when filtered with a limited number of filters) are of equal sizes. The obtained results demonstrate that the model of modules can describe to a first approximation the mechanisms performing the estimation of the sizes of the images and their segmentation. Correspondence between the data of neurophysiological, psychophysical, and modeling investigations was shown for the first time to our knowledge. The conclusions were supported by a comparison of the experimental results, and the modeling, with patterns of traditional Byzatntine icon paintings, as well as Russian avant-garde paintings of the early 20th century, which were inspired by the former. Practical significance. Application and further enhancement of the model of modules as an artificial neural network that ensures the segmentation, image size estimation, and recognition of visual objects can be of practical significance.
Read full abstract