Oscillatory neural network for adaptive dynamical image processing

We develop a biologically motivated oscillatory network model and related dynamical synchronizationbased method of image segmentation. The first version of successive segmentation algorithm was based on coupling adaptation in the oscillatory network. New model developments, presented in the paper, include: 1) a modified version of single oscillator dynamics; 2) new network connectivity rule. These modifcations permit to significantly improve the oscillatory method capabilities, providing image processing with significantly larger pixel array sizes and ensuring higher segmentation accuracy. In addition the improved network model allows to perform selective image segmentation tasks (extraction of prescribed subset of image fragments). New method capabilities have been demonstrated in computer experiments.

• Type: Conference
• Year: 2005
• Where: CIMCA
• Authors: Margarita Kuzmina, Edward A. Manykin
• Keywords: CIMCA 2005, Image Segmentation, Oscillatory Network, Oscillatory Network Model