Study On Chaotic Neural Network Algorithms And Its Application In Medical Image Processing

With the development of medical information technology, more and more medical imaging equipments have been loaded and applied in medical service organizations, and as a result there are a large volume of medical imaging data being produced, which poses challenges to effective data storage and information transmission. Therefore, it is necessary to use effective measures to guarantee data privacy and data preciseness, while to compress the medical imaging data for saving storage space. In this dissertation, some problems about medical image based on chaotic maps are studied. The main works of this dissertation are as follows.The round keys of DES algorithm are generated by two-dimensional Logistic map and two-dimensional Henon map, which can meet the requirements of dynamic changes. The improved DES algorithm is applied in medical image encryption, and the simulation results show that the encrypted image has the information with the form of random noise characteristic. The information is difficult to be deciphered by attacker, thus the image data can be protected effectively in transmission. At last, we also use several attacks to analyze the security of the improved algorithm.The diagonal recurrent neural network based on Logistic map is proposed, and its structure is designed. Then the medical image fusion algorithms based on diagonal recurrent neural network and chaotic diagonal recurrent neural network are proposed, and the fusion performance is evaluated. The fuzzy chaotic neural network is proposed. Furthermore, the structure of fuzzy chaotic neural network is designed, its mathematical model and weight training formula are derived, and the steps of image fusion are given.Two types of feed forward chaotic neural network based on Chebyshev map are proposed:the adjoin neuron of chaotic map is added to neuron of hidden layer in I type network and II type network is used chaotic map to solve the initial searching value. Then they are used in medical image compression. Experiment results show that the two types of feed forward chaotic neural network are superior to BP neural network in training time and the effect of image reconstruction. The fractal compression coding scheme based on diamond search is proposed, which aims to positioning quickly. With the use of search pattern and search path of diamond search, the goal of fast positioning is realized.