Blind motion-blur parameter estimation using edge detectors

Abstract

A method to determine the parameters of directional motion blurring noise using edge detectors is developed in this paper. Once the parameters characterizing the blurring noise are known, the point spread function (PSF) of the blurring noise can be reconstructed and the noise subsequently eliminated. The algorithm uses an edge detector with dynamic thresholding to extract edge information from an image at several threshold values. First the blurring angle is determined by rotating the image prior to edge extraction and the angle with the highest difference in edge count and its corresponding perpendicular angle is determined and considered to be the angle at which shifting has occurred. After the blurring shifting angle has been found, the pixel displacement length is determined using an iterative blind deconvolution process. Finally, once both the shifting angle and pixel displacement length are known the point spread function (PSF) of the blurring noise corrupting the image can be reconstituted.