Texture segmentation is a very difficult process in image processing and using conventional methods doesn't result in proper segmentation. In this paper we presented a novel approach in texture analysis by using neural networks. We used a special type of neuron-a multivalued neuron. The inputs in the network are spectral coefficients (DFT) with maximal values. The achieved results, applied in the processing of microscopic images of mechanical materials, indicate that this is a right way to improve segmentation of very complex digital images.
展开▼
