Abstract: In this study, an improved version of K-means algorithm is proposed efficiently to segment the satellite images. Segmentation of image is one of the promising and active researches in recent year. As literature prove that region segmentation will produce better results. Human visual perception is more efficient than any machine vision systems for extracting semantic information from an image. A K-means algorithm is developed to estimate parameters of the prior probabilities and likelihood probabilities. So, K-means algorithm is used for segmenting background and island extraction is done based on pixel intensity. Finally, Peak Signal to Noise Ratio (PSNR) is calculated and it has better results than other.