A new image restoration algorithm is proposed by combining the Electromagnetism-like Mechanism algorithm with image restoration, because reasons of degradation are very complex and methods used commonly are limited by prior knowledge. First, the degraded image is restored with Lucy-Richardson method, and random noise is added to difierent areas of the restored image to generate the initial population (i.e. the speculated image). Second, the speculated image is used to generate another degraded image by the same degradation. Then, the rational function of these two degraded images is taken to be the objective function to get the optimal restored image by seeking for the maximum value of the objective function. Experimental results show that the image restoration algorithm proposed is good at image restoration and not limited by prior knowledge.