Abstract: Color image region segmentation has a significant contribution in analysis of images and retrieval of relevant images. Color image region segmentation helps the end user to sub-divides the user input images into unique homogenous regions of similar pixels based on pixel property. It is the high end image description of objects, scenes and features. The success of image analysis mainly focused on segmentation reliability. Automatic sub-division of the user input images into unique homogenous accurate regions without over-segmentation is a tedious task. Our study is focused to segment color images automatically into multiple regions accurately based on local maxima of GLCM texture property and pixels are spatially clustered into identical regions. In the proposed method, for the given input image, the number of regions R is identified based on image texture property. Then, the images are clustered into distinct R regions which are based on median centroid. Our proposed approach generated reliable, accurate and non-overlapped multipleregions for the given user input image. Segmented regions can be automatically annotated with distinct labels which in turn help to retrieve relevant images based on image semantics.
A. Kalaivani and S. Chitrakala, 2016. Robust Automatic Color Image Region Segmentation. Asian Journal of Information Technology, 15: 5028-5037.