CCTA-based region-wise segmentation

Unsupervised image segmentation is an important and difficult technique in pattern recognition. In this paper, we propose an interesting region merging algorithm for segmentation of natural images. It consists of two steps: first forming initial over-segmentation by the Connected Coherence Tree Algorithm (CCTA), and then merging the primitive regions in terms of their similarity and feature in the spatial domain. Extensive experiments and comparisons are conducted on a wide variety of natural images. Results are good even on these complex images.