Biologically Motivated Oscillatory Network Model for Dynamical Image Segmentation

We continue to develop the 2D oscillatory network model with self-organized dynamical coupling and synchronizationbased performance for image segmentation tasks. The model has been extracted via proper reduction from 3D oscillatory network, previously designed by the authors as oscillatory model of the brain visual cortex (VC). Known neurobiological data on VC structure and functioning were re ected in design of the 3D network architecture and single oscillator dynamics, and the idea of dynamical binding was attracted in construction of network connectivity rule. Under supplemented coupling adaptation method the 2D network is capable to provide accurate segmentation of gray-level brightness images and to perform some texture segmentation tasks. Besides, the network demonstrates capabilities of smooth contour integration. New advanced versions of internal oscillator dynamics and network connectivity rule have been designed and introduced into the 2D model at current study. The modi ed network characteristics encompass promising possibilities of more e cient network performance, providing higher accuracy of brightness image segmentation.