Based on the development of digital neutron imaging system, the restoration algorithm was presented to remedy the geometric unsharpness of digital neutron radiation images. Firstly, the collimation component of digital neutron imaging system was analyzed, and the point spread function (PSF) which results in the geometric unsharpness of the neutron radiation image was estimated. Hereby, a regularized Lucy-Richardson(LR) algorithm was proposed, the algorithm investigated the bivariate model of wavelet coefficients using Bayesian maximum a posteriori(MAP) estimation theory to derive an efficient denoising method, and incorporated the wavelet denoising into the iterative step of original LR algorithm to solve the drawback of noise amplification. A real neutron radiation image of the specimen under test has been restored using this reformed algorithm. Experiments show that the restoration results are improved significantly compared with the results obtained by using the traditional LR algorithm and Fourier-Wavelet regularized algorithm. This method can also be applied to other image restoration problems.