Self-organizing feature map and linear discriminant analysis based image segmentation

This paper presents an image segmentation of images by unsupervised clustering method based on self organizing feature map (SOFM) and linear discriminant analysis (LDA). The SOFM is used for initial segmentation of the images in unsupervised method. Subsequently, the sampled image pixels of the segmented image are used for better segmented results through linear discriminant analysis. The performance of this method for segmentation is then compared for evaluation with other methods using a Davies-Bouldin index (DB-index) measure, and is found to yield comparable results for a set of natural images.