Image segmentation based on rough set theory and neural networks

A new method for image segmentation based on rough set theory and neural networks is proposed. First, the rough set is used to reduce the image attributes, extract rules, draw out the key components of image as the input of the neural networks; then, it ascertains the number of neurons in the hidden layers according to the rules and revises the weight of the neural networks by the attribute essentiality of rough set theory. Experimental results show that the method has a greater ability on resisting noise. At the same time, it solves the problem that happens in image segmentation by only using neural networks, such as “blind spots” of the neurons, the complicated structure of the networks, slower speed of constringency and so on. It greatly shortens the training time of the networks while improving the result of segmentation.