Image registration is matching and adding two or several images obtained bydifferent time, different sensors, or different conditions(weather, intensity of illumination,camera position, etc.). And it has wide applications in many fields, such as medical imageprocessing, remote sensing image-process, computer vision, etc. Also it is very importantin image mosaic, target recognition, change detection, image fusion, and timing imageanalysis. Especially, MRI image registration has great helpful for the MRI timing imageanalysis, epilepsy diagnosis. The intersecting cortical model(ICM) known as the thirdgeneration of artificial neural network, because of the characterization, no training and nolearning, compared with traditional neural networks, is already been used widely in imageprocessing, such as the image smooth, image denoising, image segmentation, et al. And inthis thesis, we will use the ICM in MRI image registration.Firstly, this thesis introduces the development of the technology of image registrationand some methods of image registration; and then expound the principle of the PulseCoupled Neural Network(PCNN) and its simplification version: intersecting corticalmodel. Also recount the mechanism in image processing application. Meanwhile, pointout the problems of the common methods of ICM: time series and time matrix, which isalready used in image process. Based on this, in this thesis, we do certain improvement ofthe ICM and put forward the method which is the Firing Time Matrix. And use thismethod in image registration and the experiment results show good effect. In order toreduce the running time of the ICM used in image process, we use the edge detectionmethod to lessen the convolution operation of the model. So this thesis mainly contains:1. Introduce image registration technology: principle, classification and evaluation,etc.;2. Research the ICM, understand its advantages and disadvantages, and then putforward the Firing Time Matrix method;3. Use edge detection method to reduce running time of the model applied in imageprocess. |