A new ensemble clustering method for PolSAR image segmentation

In this paper, an effort is made to integrate spectral clustering and Gabor feature clustering, leading to improved segmentation results. The spectral clustering divides an image into nonoverlapped groups such that the intragroup similarity is high and the intergroup similarity is low as much as possible. This method includes solving the eigenvalue problem for the normalized similarity matrix, of size n × n, where n is the number of pixels. On the other hand, Gabor filter is used for texture feature extraction. A texture feature vector for each pixel of the image is formed corresponding to the texture edge energy at different directions with Gabor filter. The K-means clustering is applied on the texture feature vectors of all pixel of the input image. Finally, to integrate the results of spectral clustering and Gabor feature clustering, a cluster ensemble approach is applied and PolSAR image segmentation is performed. The experimental results indicate the effect of proposed method on PolSAR image segmentation.