Journal of Electronic Systems Ken’ichi Fujimoto, Mio Kobayashi, Tetsuya Yoshinaga 3 1 2013 http://www.dline.info/jes/fulltext/v3n1/3.pdf Image segmentation is a fundamental technique in image processing systems such as a medical computer-aideddiagnosis system. We previously proposed a discrete-time oscillator network system with a degradation (posterization) system for dynamic image segmentation. This system can extract image regions with similar pixel values (connected components) from a given gray-scale image and exhibit extracted image regions in a time series. The dynamics of the proposed system is described by ordinary difference equations. In spite of discrete-time system, our oscillator model can generate oscillatory responses like relaxation oscillations observed in ordinary differential equations. Dynamic image segmentation for a gray-scale image is performed according to posterization of pixel values and the synchronization of oscillatory responses from discrete-time oscillators. Since numerical integration is unnecessary to work the proposed system, its processing speed is significantly faster than that of another dynamic image-segmentation system consisting of ordinary differential equations. In this paper, for further speed-up of dynamic image-segmentation processing, we presented an implementation of the proposed system into a graphics processing unit.