Image Processing via Synchronization in a Self-Organizing Oscillatory Network

Abstract

We develop a biologically inspired method of image processing based on synchronization-based performance of an oscillatory network with controllable self-organized coupling. The oscillatory network, obtained from a previously designed biologically motivated oscillatory neural network model of the brain’s visual cortex, provides automatic, adaptive, and active image segmentation. Being tuned by an image to be processed, the network dynamics realizes network decomposition into a set of synchronized ensembles of oscillators, corresponding to image decomposition into the required set of image fragments. The current network model version provides: (a) full segmentation of real grey-level and colored images; and (b) selective image segmentation (extraction of a subset of image fragments with brightness values contained inside a given arbitrary brightness interval).