Unsupervised segmentation of large scale spatial images using K-means clustering approach

Segmentation is one of the most critical means of image processing and data analysis approach. The aim of segmentation is to classify an image into parts that have a strong correlation with objects in order to reflect the actual information collected from the real world. Clustering is a major global approach for segmentation and it relies on partition of images into a set of layers or regions for further analysis. The image segmentation by clustering basically refers to grouping similar data points into different clusters. In this article, an unsupervised clustering technology is proposed for processing large scale satellite images taken from remote celestial sites where none explicit teacher is introduced. As an effective approach, K-means clustering method requires that certain number of clusters for partitioning be specified and its distance metric be defined to quantify relative orientation of objects. Then image processing system forms clusters from input patterns. Diversified large scale image features are investigated using unsupervised methods. In the mean while, to limit computational complexity for the consideration of real time processing, a simple study is conducted where tristimulus values are selected to represent three-layer color space. Simulation results show that this approach is very successful for spatial image processing.