This paper presented a novel image zooming algorithm based on maximum a posteriori(MAP)probability theory and maximum entropy principle.Compared with traditional Tikhonov regularization linear method,the proposed algorithm can deal with the noise removal and pixel interpolation using the gradient information self adaptively,which is used as an anisotropic filter.Both theory analysis and experimental results show that using this method,image edges can be preserved while the disturbed problem of noise points are also conquered,our method is effective for image zooming.