Image Restoration Using Genetic Algorithms

Video shooting conditions are almost never close to ideal. Even under favorable conditions (good lighting, diffused light, etc.), the resulting image may be blurred because of camera movement or poor focus. This problem is especially acute in fields where image restoration can be very expensive or even impossible. There are many methods to solve this problem; however, none of them completely solves it. The existing approaches can be divided into two classes: manual restoration and automatic restoration. For manual restoration, it is necessary to have information about the factors that caused blurring, as well as about their parameters. In practice, these parameters are almost always not known. Therefore, the problem of automatic restoration of blurred images (blind deconvolution) is of greater interest and importance. The process of solving this problem in blind deconvolution methods consists in optimizing a certain function. This paper presents an overview of existing image restoration methods and proposes a restoration procedure based on an optimization model that uses genetic algorithms with certain recombination operators. For optimization, blind deconvolution mainly uses gradient methods. However, these methods have a flaw: they tend to stuck in local minima. Genetic algorithms are more robust. However, despite this advantage, there are few studies devoted to the use of the genetic algorithm as an optimization model for blind deconvolution. The results of our research, including some specific aspects of using genetic algorithms, are discussed at the end of the paper.