ブラインド・デコンボリューションに遺伝的アルゴリズムを応用した画像復元法

概要

論文の詳細を見る
This paper describes a method for image restoration by blind deconvolution suing genetic algorithm (GA), called Genetic-Iterative Fourier Transformation Algorithm (GIFTA). Firstly, an image which is guessed by a given blurred image, called initial guess image, is generated. And the initial guess image is considered to be an original image. Then, the blurred image and the initial guess image are Fourier transformed. Because an intensity of Fourier transformed image is defined as a gene. Secondly, an initial population is made by a phase of blurred image and one of the two intensity. The intensity is selected at random. Then, the gene is estimated by simularity of images. So the population is inverse Fourier transformed for estimating the gene. Where, a-priori knowledge which is non-negativity and finitly of an image is considered when inverse Fourier transformation is operating. Thirdly, the population is Fourier transformed once again for genetic operation, i.e., reproduction, crossover and mutation. And these steps are iterated until ending condition. Finally, experimental results using this method are presented. And the convergence of the algorithm is confirmed.
東海大学の論文