Piecewise linear motion blur identification using morphological filtering in frequency domain

A new method for extracting blur parameters in piecewise linear uniform motion blur is proposed. In the case of piecewise linear motion blur, we should know both the blur parameters and the order of their combination. The blur parameters, which consist of more than two directions and lengths of motion, can be extracted from the frequency response of the blurred image. The morphological filtering is employed to enhance the precision of the directions and the lengths. Further, the estimated point spread function (PSF) of the motion blur is selected from possible PSFs, which is obtained by combining the extracted parameters randomly, by finding the one which has the maximum value of correlation with the degraded image. The experimental result shows the performance of the algorithm.