Extracting nonlinear features for multispectral images by FCMC and KPCA

Classification is a very important task for scene interpretation and other applications of multispectral images. Feature extraction is a key step for classification. By extracting more nonlinear features than corresponding number of linear features in original feature space, classification accuracy for multispectral images can be improved greatly. Therefore, in this paper, an approach based on the fuzzy c-means clustering (FCMC) and kernel principal component analysis (KPCA) is proposed to resolve the problem of multispectral images. The main contribution of this paper is to provide a good preprocessed method for classifying these images. Finally, some experimental results demonstrate that our proposed method is effective and efficient for analyzing the multispectral images.  2004 Elsevier Inc. All rights reserved.