A Research On The Models Based On The Biological Characters Of Human Vision System And Applications

As one of human's important organs and the first way to acquire and deal with information (specially the information about images), eyes have more superior biological mechanisms than those in normal optical ways obviously. The main work of this paper is not only to build and analyze the mathematic models of human vision system based on the features of human vision but also to prove the realization of those models and apply them to the medical image processing. It has improved the results of images processed obviously. The research work is focused on the following several aspects:(1) Through the mechanism called lateral inhibitory existing among the visual neurons, visual information can be selectively detected. Simulating this mechanism among visual cellars through a mathematical method, a shunting inhibitory retinal cellular neural networks(SIRCNN) model based on early vision is provided. The stability analysis of its dynamics and the equivalent circuit diagram are presented in this paper.(2) Based on SIRCNN, a new algorithm of detecting edges is provided. In terms of the features of receptive field of vision cells, suitable forward templates are generated. From the test's result, such method based on biological visual mechanisms cans efficiently enhancing and detecting edges.(3) Through the mechanism called locally excitatory globally inhibitory existing among the visual cortical neurons, neurons can oscillate in terms of visual information. In this paper, the model of locally excitatory globally inhibitory visual cortical neuronal oscillator network with the mechanism among visual neurons is assayed and studied deeply. An analytic solution to nonlinear dynamic system of this model is obtained in this paper by using disturbing method.(4) Based on the model of locally excitatory globally inhibitory visual cortical neuronal oscillator network with those mechanisms among visual neurons and the theory of cross entropy, an adaptive parameter setting method of image segmentation is provided and then applied to the medical image processing. From the test's result, such method based on biological visual mechanisms can efficiently detect different targets.(5) Based on the researches of the two models, the relationships and differences between them has been assayed deeply in this paper in three aspects : physiologicalmechanisms, mathematic theories and mechanisms of vision information processing.