A0320
Title: Unsupervised fuzzy statistical learning and its applications in image segmentation
Authors: Siu Kai Choy - The Hang Seng University of Hong Kong (Hong Kong) [presenting]
Yee Lam Mo - The Hang Seng University of Hong Kong (Hong Kong)
Abstract: Fuzzy clustering algorithms, statistical modelling and spatial statistics are popular methodologies in image processing and pattern recognition. However, the literature has not studied the integration of these techniques in image segmentation applications. A robust and effective fuzzy-model-based unsupervised learning algorithm is presented that integrates colour and the wavelet-domain generalized Gaussian density (GGD) statistical model with the fuzzy clustering algorithm combined with neighbouring information for image segmentation applications. Using the GGD statistical model to characterize wavelet subband texture information, the proposed algorithm is particularly effective in segmenting images with complex texture patterns. Comparative experimental results with current existing fuzzy clustering-based approaches show that this methodology achieves remarkable success in image segmentation applications.
