Dissertation > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device

New Methods of Color Image Segmentation Based on Neural Network Model

Author ChenYuZuo
Tutor FanZhouTian
School Beijing University of Technology
Course Mathematics
Keywords neural network model stability analysis image segmentation synchronous power spectrum
CLC TP391.41
Type Master's thesis
Year 2013
Downloads 4
Quotes 0
Download Dissertation

Neural network models are of great significant in image processing.In this paper, two network models are developed for color image segmentationfirstly, we use the network based on Wilson-Cowan model to implementsynchronization because of the similar external input and local coupling. Limit circleis used to simulate the activity of neuron, however, the conditions for generating limitcircle is changing with the external input, the segmentation result is not ideal. ThenKuramoto phase model is taken into consideration, and the periodic activity of neuronoscillator is described completely. New mapping system is established based onKuramoto phase model, using the harmonic spectrum analysis and spectrum principle,the activity curve of neuronal oscillator and the instantaneous frequency variationcurve are depicted. Combining the knowledge of power spectrum and the theory ofinstantaneous frequency synchronization, color image segmentation is done, and thesegmentation results are compared with other results from approaches.This approach has the advantage of adjusting segmentation region and strengthby changing the threshold of instantaneous frequency, and the algorithm out-performssome other algorithms from other approaches.

Related Dissertations
More Dissertations