Abstract: To improve the segmentation effect of complex geometric transformation or disturbing images, a novel shape descriptor and texture segmentation algorithm is proposed. Image statistics are aggregated in the region of interest, shape descriptors are trained by neural network, and texture segmentation is carried out by using the trained shape descriptors. The segmentation experiments of synthetic image and real image show that the proposed algorithm is superior to other algorithms in contour index and region index. Experimental results show that the proposed algorithm is effective and feasible, and can achieve a better segmentation effect for complex geometric transformation or disturbing images.