Abstract:
A image segmentation algorithm based on improved spectral clustering and particle swarm optimization is proposed.Similarity matrix is constructed by the mean of dual -tree complex wavelet transform coefficients in this dissertation so as to make full use of the spatial adjacency information and feature similarity information included in the data.To efficiently apply the algorithm to image segmentation,Nystr?m approximation strategy is used in the course of spectral mapping to reduce the computation complexity and memory consumption.And then we tentatively adopt particle swarm optimization algorithm to optimize the K-means clustering in the spectral clustering algorithm.Experimental results on medical images and remote sensing images verify the validity of the proposed algorithm.