Hough transform and cepstrum based estimation of spatial-invariant and variant motion blur parameters

Motion blur is a most common degradation present in the photograph captured from hand held camera. Minute visual details are lost and spreaded over pixels creating smearing impact in captured image due to motion blur. It's hazardous for machine vision application and should be corrected by any restoration methodology. As no information is available about kind of degradation and true image, restoration of single degraded image is considered ill-posed inverse problem. In this paper simple and efficient technique is given for correct estimation of motion blur parameters: blur direction and blur length. Hough transform of the fourth bit plane of the modified cepstrum is employed for blur direction estimation under both spatial-invariant and variant blur consideration. Grayscale transformed modified cepstrum is utilized for blur length detection. Experimental results on standard images for wide range of angles and lengths support the usability and accuracy of the proposed method. Compare to already existing strategies proposed method estimate correct values even if noise is present. These estimated motion blur parameters values are than applied in parametric wiener filter based deconvolution algorithm to get restored image.