Fuzzy Genetic Algorithm And Its Application On The Image Restoration

An image is a kind of significant information that the human acquires. A large number of sharp images play an important part in our daily life and scientific researches. However, an image is always influenced by many factors in the process of creation, transmission and recording. Thus, usually, an image gotten by different ways is not with intact information but a loss one. Therefore, scientific researches on image restoration become more and more imperative.Because of the complexity and close correlation of the image information, at different levels in image processing maybe some problems arise, such as non-integrity, non-precision and non-structure. In conjunction with computational intelligence methods, image restoration can get a better effect in comparison with the traditional methods.The genetic algorithm (GA) as an intelligent optimization algorithm makes use of the natural selecting and genetic rules for stochastic searching. The primary character of GA is that it can operate at multiple points in the searching space synchronously for the global optimization. The way of searching is led by the stochastic rules and is independent of the gradient information. So GA is capable of solving the nonlinear problems which may cause difficulty for the traditional methods. On the other side, the fuzzy control theory emphasizes the macroscopical function described with certain language or conceptions. Since it is close to the human intellection and cognition, the fuzzy control theory can express effectively the recognition activity or other complicated things. Additionally, it has the ability of control and guidance, which can solve the complicated control problems easily. The image restoration based on the genetic algorithm and fuzzy logic control using allowed non-integrity, non-precision and partial facticity of information can find a solution which is easy to be processed, robust and cost-effective. Consequently, this method has great potentialities for further development of image restoration technology.This thesis focuses upon the improvement of the fuzzy genetic algorithm and its application on image restoration, mainly the following issues.First, aiming at the premature convergence of the genetic algorithm, an improved fuzzy genetic algorithm is proposed. In this algorithm, the mean square deviation of group fitness and population evolution generation are used as the criteria of premature convergence, and according to the estimation from fuzzy logic controllers, relevant evolution methods are given to different chromosomes, that is punishing the strongers and awarding the weakers when the population evolves normally in order to maintain the diversity of population, while doing catastrophe operation to the weakers to renew population evolution once the premature convergence appears or tends to appear. The experiment results show that the improved fuzzy genetic algorithm can maintain the population diversity and suppress the premature convergence better in comparison with other kinds of the genetic algorithms.Then, in order to avoid the high computational complexity of the genetic algorithm for image restoration, a kind of blocking restoration method based on the fuzzy genetic algorithm is proposed. A strategy of "divide and rule" is adopted, which divides averagely a blurred gray image into blocks and utilizes the genetic algorithm to restore them sequentially. Since deblurring blocks respectively can lessen the quantity of data during the genetic evolutionary process, every block can obtain preferable effects in a short time. In addition, the gray-level distribution of each block is diverse, some blocks are simple backgrounds with unitary gray levels, some are details with wide-ranging distributions, the others may be the borderlands between the two instances, having strong edge information. Hence the fuzzy logic controller is adopted to tune the crucial parameters of the genetic algorithm for better recovery performance, including the parameters of fitness function, the initial temperature of simulated annealing for fitness scaling, the probabilities of crossover and mutation. Finally, the recovered blocks are recombined to a whole image, and the boundary errors caused by block processing are damped in the meantime. Experiment results show that the presented method has larger predominance on the operation speed, the memory cost and the restoration quality than the conventional genetic algorithm.Finally, the fuzzy genetic algorithm is used for blind image restoration. In this thesis, the point spread function (PSF) is supposed to the space variant, and its type is unknown or partially determined. As the above blocking method brings on serious boundary noise, the image blocks are divided by the triangle meshes according to the image degradation. With a view to the correlation between the adjacent points, the neighboring triangles with similar kurtosis comprise an image block according to the 'Central Limit Theorem'. On the other hand, the micro genetic algorithm (micro-GA) is adopted with the standard genetic algorithm (SGA) to alternately estimate the image block and the corresponding point spread function. The micro-GA has small scale of population which lowers computation complexity while performing the fine tuning in the overall solution. These advantages compensate the limitation of SGA for image restoration. When the type of PSF is known, the parameter of PSF is estimated by the SGA, otherwise the PSF matrix is estimated. Especially, when the type of PSF is unknown, the best estimated PSF is corrected by space invariant PSF models based on the fuzzy controller at each generation of SGA. The amending weights are the similarity between the estimated PSF and the known models. In addition, to enhance the power of removing noise and protecting the details, the image block is also mended according to the histogram statistics during the iterations of micro-GA. Experiment results show that the presented method can restore effectively different space variant blurred images, and its power of suppressing noise is also strong.