Neural network based SOM for multispectral image segmentation in RGB and HSV color space

Segmentation is the process of partitioning an image into number of meaningful images as segments or clusters. The segmentation is initial but important process which is used to locate boundaries and objects in images. This paper is concerned with segmentation of color satellite images using neural network based kohonen's self-organizing maps. This unsupervised competitive network is used to visualize and interpret large data sets. In this paper, test images are segmented in RGB and HSV color space using self-organizing map and the segmentation results are compared using error image, peak signal to noise ratio, and execution time. The efficiency of proposed method is tested with Landsat and Terra (MODIS sensor) satellite images.