Structured Mean Shift Method and Its Application on Image Segmentation

It was found that the traditional Mean Shift algorithm segmentation often produces the connection channel, several categories between clusters can not be completely separated from the problem, an improved classification of the structured mean shift segmentation algorithm iterations to obtain the initial cluster in center on the basis of different bandwidth matrix stepwise clustering, resulting in a different class collection of cluster centers, and the establishment of a vested in the tree structure, ultimately through the attribution of the leaf nodes and the root node are classified in order to complete the image segmentation. Experiments show that the algorithm can better retain the image of local information, and has better applicability.