Histogram based hill climbing optimization for the segmentation of region of interest in satellite images

Images received from the satellite contains huge amount of information to process and analyze. So the segmentation is a crucial and important procedure in the analysis of images to gather necessary information from the satellite images. In the proposed approach, the satellite images are segmented using hill climbing local optimization technique and modified k-means clustering algorithm. In this approach, satellite images in RGB color space is converted into CIELAB color space. This color space is intended to approximate vision of human and perceptually uniform. Moreover, the intensity (L) component of this color space exactly matches the human perception of lightness. In the next step, the hill climbing process is applied on the color histogram of CIELAB color space image to obtain the initial cluster centers. In the final step, these cluster centers are given to the k-means clustering algorithm to produce the segmented image as the output. The effectiveness of the proposed approach has been demonstrated by number of experiments. The proposed method is more effective and efficient in the segmentation of satellite images to obtain meaningful clusters as compared to other conventional methods.