Image Restoration Based On Intelligent Algorithm

Image Restoration is a task that using priori knowledge from degradation image and build a mathematic model which can describe the degradation phenomenon. According to this model, we obtain an original image by inversion. Image Restoration is a core issue in Image processing and computer vision. Genetic algorithm and artificial neural network are two Intelligent Optimization Algorithms. There are lots of practical application is people's daily life. Nowadays the dimensions of image became larger and lager, with this we need better algorithm which can solve the problem effective and faster. Single algorithm has its own shortcomings that can't solve the problem, so we think that combine two or more better algorithm may overcome the limitation of single method and then obtain better restoration results. This article mainly contains three parts based on the idea:1.Genetic algorithm is one of the most important search methods in intelligent search areas. Because of its own advantage that it has no extra restrictions, GA can change the problem into environment and solve the problem. GA is a parallel algorithm and has good robustness, so GA can apply in image restoration and solve the problem. Original GA has its own weakness like premature convergence. This article focus on this point introduce some solutions can solve this problem and we put forward a new improved method based on chaos adaptive, new algorithm has better results according to our experiment.2.Original Image Restoration algorithms have limitations, they can't meet the need of real-time processing. Along with Artificial neural network develop, a new algorithm in Image Restoration has been put forward. Because of ANN's own advantage like distributed memory and parallel processing, Image Restoration based on ANN can solve the problem effectively and faster. This article give some improved Hopfield algorithms and improved RBF network, experiments can prove the methods are better than original methods.3.In order to overcome the limitation of single algorithm, article put forward a new method which combines Hopfield NN and GA. To some extent new algorithm improved the single method and obtains better results. And then this article combines Chaos Adaptive GA and Hopfield, this new method improved original single method and GA+Hopfield NN. In order to overcome the premature convergence of GA, we use improved GA like Chaos Adaptive GA to improved RBF neural network, our experiment shows that we obtain better results when compare new algorithm and original algorithm, and can lower compute cost.