Image Segmentation Using Intersecting Cortical Model

Image segmentation has been a very active research field in digital image processing since 1950s. As an important image processing technology, it plays a key role in advance from image processing to image analyzing and has been paid more attention in research and application, because the accuracy of image segmentation affects the efficiency of later works directly.The paper first analyzes the numerous research results in the field of image segmentation, as well as the practical applications and the hot issues of current research of image segmentation. And give a brief introduction to the application of pulse coupled neuron networks (PCNN) in digital image processing. After studying the PCNN models we find that it is not fit for real-time processing, so in order to effectively overcome the difficulties encountered in PCNN based image segmentation, the intersecting cortical model (ICM), proposed by Kinser et al, is employed in this paper. ICM is a kind of new biological visual model and a hot spot of current biological visual models. With the biological background, ICM was designed according to models of the brain visual cortex of many kinds of mammalian and based on the public portion of them, so ICM suits the real time processing of images and graphs without loss of the basic acting mechanism of the model of visual cortex neuron. Then the structure, mechanism and properties of ICM are elaborated when compared to PCNN, and is used to achieve automatic segmentation of images. The deep study of its application to image segmentation is also conducted.It is a key technical problem to select criteria and determine the optimal segmentation results for image automatic segmentation under ICM. This paper proposes a new criterion named simple variation. A large number of experiments show that this algorithm is better than other classical algorithms such as cross entropy, region homogeneity, region contrast. Under the condition of the lossless of segmentation results, the ICM algorithms is effective implementation, and the experimental results show that ICM is more satisfied for the real-time processing.