The accurate estimation of point spread function (PSF) is the key to the image degradation model in motion blurred image restoration. An algorithm to calculate the PSF of blurring caused by uniform linear motion in arbitrary direction is presented. By computing the extent of superposition between neighboring pixels along the motion direction, the two-dimensional PSF is estimated and then used in image deconvolution to remove the motion blur. We implemented both computer simulation and imaging experiments to demonstrate the proposed approach. The image restoration performance is evaluated by comparing our method with existing algorithms via image quality metric functions.