A Normalized Cuts Based Image Segmentation Method

Abstract

A new image segmentation method is proposed in the framework of Normalized Cuts to solve the perceptual grouping problem by means of graph partitioning, and the multiscale graph decomposition to obtain image features. Texture features is modeled with orientation histograms defined on the different scale level. The global optimal segmentation can be efficiently computed via graph cuts. The segmentation is implemented by partitioning a graph representing an image at the finest scale level to obtain accurate segmentation, while the weights of the graph are calculated from all the scales. Due to the reduced dimensionality based on texton, the speed of Normalized Cuts is increased. Efficiency and accuracy of the method is demonstrated on the nature images and remote sensing images segmentation.