Finding point spread function of motion blur using radon transform and modeling the motion length

Abstract

The success of restoring images degraded by motion blur highly depends on precise estimation of parameters such as motion direction and length that were involved in the motion blur function. In this paper we suggest a method which uses Radon transform to estimate motion direction and for estimating motion length we used curve fitting methods to provide an equation that could model the behaviour of Fourier transform of images that were degraded with different motion lengths. The input to this model is a property found to be available in the Fourier transform of the degraded image and its output is an estimate for the motion length. Our method was tested on a wide range of different type of images that were degraded with different direction and motion lengths. The low values of mean and standard deviation of errors in our estimations, showed the preciseness of our method.