The neural network method research of color image used in micro stereo measuring

Micro stereovision based on stereo light microscope (SLM) is used in Micro stereo measurement, in which stereo matching is an important part. In this paper, a new method for solving this problem is proposed by raising an optimized stereo matching algorithm based on Hopfield Neural network and color stereo image pair getting from SLM. In this method we firstly define the color similarity by using the internal association of RGB in a color image, and then combine it with Rank transform and the NCC algorithm to build the optimal stereo matching relations. At last, we use these relations to build the Energy function of Hopfield network combined with the inherent constraints of computer stereo vision. The experimental results show that the accuracy of stereo matching and noise immunity are partially improved by using Hopfield network algorithm compared to NCC algorithm.