An adaptive clustering and chrominance-based merging approach for image segmentation and abstraction

Abstract

We present a novel, computationally efficient approach for natural image segmentation that uses the adaptive clustering algorithm (ACA) to obtain an initial segmentation and chrominance-based region merging to consolidate regions of perceptually uniform texture. The combination of ACA and chrominance-based merging preserves salient edges and smooths out noise and edges within textured regions. It can thus be used for image abstraction. Experimental results with natural images indicate the effectiveness of the proposed approach.