International Symposium on Nonlinear Theory and its Applications
Extraction of Image Regions Using Oscillatory Responses in Chaotic Neuronal Network
Mio Musashi, Ken’ichi Fujimoto, Tetsuya Yoshinaga,
PDF download (381.7KB)
For dynamic image segmentation, which is used to extract connected components in an image and to exhibit them in a time series, we have developed a network model consisting of discrete-time chaotic neurons. According to an n-phase oscillatory response of neurons, the network model can segment n image regions within one processing. Based on the bifurcation analysis of periodic points in a reduced model of the network, we propose a successive algorithm for dynamic image segmentation of a gray-level image with an arbitrary number of isolated image regions, using our network model with two- and three- phase oscillatory responses.