On improving accuracy of PSF estimation in spectral and cepstrum domain with morphological filtering

An improved motion blur point spread function estimation method for a single uniform motion blurred image is proposed in this paper. In the blind deblurring problem, to deblur an image it is required to estimate the blur parameters i.e., the direction and length of motion as the primary step. These parameters can be estimated from the frequency and cepstral domain representation of the blurred image. In the proposed work, the morphological filtering operations are employed to increase the precision of blur parameter estimation. These estimated parameters are used to construct the point spread function (PSF) of the motion blur. With the help of near accurate estimated blur kernel to the original one, we can observe a good quality restoration of a uniformly blurred test images. The results obtained through exhaustive experimentation prove the improved performance of the PSF estimation scheme proposed.