With respect to the problems of the fuzzy edge and difficulty in automatic segmentation of human organs during the computed tomography (CT) scanning, a local prior shape information and active contour based model is proposed. For an object whose shape is similar with the shapes in the dictionary, the prior shape in the shape dictionary is used to supervise and guide the high-level object segmentation while underlying segmentation is performed based on the image information. On the basis of the existed shape dictionary sparse representation, dictionary shapes are decomposed by mask matrix, and supplemental dictionary is generated, so the local shapes of the object can be described by the constraint of sparse shape of partial prior shapes. By the decomposition and recombination of the local shapes instead of the traditional prior shapes, shapes which are not included directly in the dictionary can be segmented, and the application range is extended. Experimental results of the segmentation experiments show that even if the edge of the object is fuzzy, the image can be recovered and segmented accurately with the proposed method, and the proposed method can be applied to medical image segmentation.