One of the key problems of restoring a degraded image from motion blur is the estimation of the unknown non-linear blur filter from a single input blurred image. Many blind deconvolution methods typically assume frequency-domain constraints on images, simplified parametric forms for the motion path during camera shake or use multiple input images with specific characteristics. This paper proposes an algorithm for removing non-linear motion blur from a single input blurred image using Genetic Algorithms (GAs), by finding the proper parameters and goal function. Also recent research in natural image statistics is exploited, which shows that photographs of natural scenes typically obey heavy-tailed distribution. The Point Spread Function entries are used as the parameters of the GA. Experiments on a wide data set of standard images
degraded with different kernels of different sizes demonstrate the efficiency of the proposed approach especially in small blur lengths compared to other algorithms with reasonable running times for a GA.