Research Of Image Edge Detection On Cellular Neural Network

Edge is one of the important characteristics of image, and it also is the element of several research areas, like computer vision and pattern identification.Cellular Neural Network (CNN) is a parallel processor. In the erea of image processing, the CNN has a broad developing space.The dissertation firstly introduces the process of edge detection based on CNN and supplies the flow chart of binary image arithmetic designed. The network parameters are used to detect the image edge. The detection results show these parameters are suitable. Meanwhile, on the basic of binary image, the detection of gray-scale image is realized by the modified eight-level arithmetic. The tetection results also show the modified arithmetic is more effective.The two simulation models for image edge detection based on CNN arithmetic and traditional arithemestic respectively are designed. Compared the two simulation results, the CNN arithemetic has several advantages: high speed parallel calculation on hardware, calculation speed is independent of image size, real-time detect image edge. Therefore, the arithmetic of CNN is an effective method.Then, the value range of CNN model is presented, the results used these values show it is correct. On this base, the value range of gray image is analysised deeply. The relation between Model value and pixel value is gained. So the self-adapted model of gray model is confirmed.Finally, the CNN arithemetic is realized by FPGA hardware in the dissertation. The realization of hardware is described in detail. Also, the experiments results show that this method is feasible.