Affinity Fusion Graph-Based Framework for Natural Image Segmentation

This paper proposes an affinity fusion graph framework to effectively connect different graphs with highly discriminating power and nonlinearity for natural image segmentation. The proposed combines adjacency-graphs kernel spectral clustering based (KSC-graphs) according a new definition named nodes of multi-scale superpixels. These are selected on better affiliation superpixels, namely subspace-preserving representation which is generated by sparse subspace pursuit. Then KSC-graph built via novel explore the nonlinear relationships among these nodes. Moreover, adjacency-graph at each scale constructed, further used update across scales, it partitioned obtain final segmentation result. Experimental results Berkeley dataset Microsoft Research Cambridge show superiority our in comparison state-of-the-art methods. code available https://github.com/Yangzhangcst/AF-graph.