Modelling the attention zones in problems involving the automatic decomposition and structural analysis of images

Луцив В.Р. Моделирование зон внимания в задачах автоматической декомпозиции и структурного анализа изображений // Оптический журнал. 2007. Т. 74. № 4. С. 59–67.

Lutsiv V.R. Modelling the attention zones in problems involving the automatic decomposition and structural analysis of images [in Russian] // Opticheskii Zhurnal. 2007. V. 74. № 4. P. 59–67.

V. R. Lutsiv, "Modelling the attention zones in problems involving the automatic decomposition and structural analysis of images," Journal of Optical Technology. 74 (4), 274-281 (2007). https://doi.org/10.1364/JOT.74.000274

This paper presents a method of analyzing images that originate in the study of the saccadic movement of the eyes published by I. Muchnik and N. Zavalishin in 1974. We have transformed the information function that they proposed, and this made it possible to fix the attention zone (AZ) on the centers of gravity (CGs) of local objects that differ from the surroundings by the average brightness. Our approach can be organically extended by taking into account features of the texture instead of differences of the local brightness. The centering of AZs on the CGs of an object makes it possible to directly recognize images subjected to an affine transformation, as is characteristic of patterns of remote surfaces projected onto the retina. An algorithm estimates the magnitude and direction of elongation of the image of the selected but still not recognized object, as a consequence of which the AZ takes the form of an ellipse that includes the object of interest and its macroscopic surroundings. This makes it possible to isolate the images of homogeneous surfaces corresponding to separate objects, each of which can later be analyzed independently by means of the structural analyzer that we developed earlier for the recognition of two-dimensional scenes. The parameters of the spatial position and elongation of each local surface can also be used to structurally analyze the scene as a whole.