Hybrid image segmentation using watersheds and adaptive region growing

We propose a hybrid image segmentation algorithm, which combines edge and region-based techniques through the morphological algorithm of watersheds. After the initial segmentation of watershed transform on the image gradient magnitude, the region adjacent graph (RAG) technique is used to construct the graph of segmented regions. We then present a graph-theoretic approach to adaptively merge the regions based on the weighted intensity difference. Experimental results obtained with some images extracted from MPEG test sequences are presented.