Research On Image Superpixel Segmentation Based On Sparse Subspace Clustering

Image segmentation is a key image analysis technology,which solves the problem of image object extraction,and makes it more efficient in image recognition and selection by forming abstract and compact image information.With the development of new theories,many new forms of combined algorithm for image segmentation have been proposed to meet the increasing segmentation requirements.On the idea of superpixel and sparse subspace clustering,this paper studies the method of image superpixel segmentation based on sparse subspace clustering,which includs superpixel feature extraction,graph construction and graph segmentation.Firstly,the theoretical basis and model of sparse subspace clustering are introduced.Secondly,the simple linear iterative clustering(SLIC)algorithm is studied.Aiming at the problem that the number of superpixel and the compact factor need to be set manually,a parameter adaptive SLIC method is proposed,which uses HSV non-uniform quantization and color neighborhood varianc similarity weighted method to achieve parameter adaptive.Experimental results show that the proposed method can generate superpixels in line with visual perception and have the effect of anti under-segmentation and over-segmentation.Finally,sparse subspace clustering and low rank subspace clustering are used to construct the similarity matrix of the graph,which is used to map the image to the graph,and the segmentation of the graph is accomplished by normalized cuts.Experiments show that the segmentation method based on the combination of superpixel and sparse subspace clustering can extract the salient object in the image.