RBFN Based Motion Blur Parameter Estimation

The image restoration techniques are oriented towards constructing the original image from a degraded observation. Motion blur in images is one of the most common degradation phenomena. Restoration of such images require correct estimation of motion blur parameters. In this paper we have proposed a novel technique for estimating the angle and the length of motion blur. Gabor filter has been used to find the angle of the blur and a simple neural network has been used to find the length of blur. The proposed method was tested on various standard images. The experimental results reveal that proposed method works better than the existing methods in terms of visual quality and PSNR.