Image segmentation via manifold spectral clustering

In this paper, we propose a novel image segmentation method based on manifold spectral clustering. This method is based on the simple idea that image can be represented as the set of several manifolds which are also referred as super-pixels, and thus image segmentation problem are solved by manifold clustering. Based on this idea, we have designed a novel manifold spectral clustering method for image segmentation. The proposed method consists of four main steps: manifold generation, manifold representation, manifold distance, and manifold clustering. Experiments are performed on many different kinds of synthetic data and natural images to verify the effectiveness of the proposed method.