Based On Cellular Neural Networks Image Processing

In this paper,we lay primary emphasis on the template design and application in image processing of cellular neural networks(CNN).Mainly we design template of CNN with something about edge detecting of image,with which we run simulations in PC(Personal Computer).At the same time,the virtual CNN which is constructed by the devices such as DSP,FPGA is used to process Real-time image.In the paper,we study and discuss key algorithms and technique about system realization.This thesis consists of the following parts. (1) This article introduces international and internal actuality of theory and application of artificial neural networks ,specially CNN. (2) The theory and the architecture of CNN are studied.Dynamic range of CNN is expounded.Stability of CNN is analyzed.Then we do computer simulation of dynamic range of CNN simply.The architechture and work procedure of chip that names ACE16K based on theory of CNN is introduced. (3) The Classification and the application of image pre-processing are analyzed.Edge detecting of image is illuminated weightily and several kinds of edge detectors are listed. (4) Templates of CNN are designed. Relaxation method is studied in its convergence,on the base of which relaxation-annealing algorithm(RAA) is put forward.The template of CNN that is designed through RAA is applied to edge detecting and well-result is gain.Then the genetic algorithms(GA) is studied and improved GA is used to designed template of CNN.The template designed through GA is applied to edge detecting of image and get right outcome. (5) Image processing system consist of DSP,FPGA etc. ,we can get virtul CNN with CNN algorithm embedded in DSP ,then the system is applied to image Real-time processing.CNN is described in VHDL and its experimental platform is constructed in FPGA.