Algorithm Study For Image Segmentation Based On Artificial Neural Network

Image Segmentation is one of the most important image analysis techniques. This thesis deals with gray image. The kernel of segmentation is pixel clustering. It belongs to optimization problem. In order to apply Hopfield neural network to optimization problem, we should translate object function into energy function and make the variables of the problem correspond to the states of the network. When the energy function converges at minimal value, the states of the network correspond to the optimal solution of the problem. It is the most difficult to determine the weight coefficients of the energy function. If we introduce the competitive mechanism, to determine the pertinent parameters is easy. In addition, we introduce the fuzzy mechanism and spacial mechanism so that the results are insensitive to noise. Experimental results show that the method of this thesis gets a better result.