Improvement of Point Spread Function Estimation Method for Motion Blurred Image and Image Restoration

Abstract

The estimation of point spread function (PSF) of motion-blurred images is the key to image restoration, which is mainly affected by two physical quantities: blur angle and blur length. The traditional point spread function estimation method is easy to be disturbed by the bright cross in the centre of the blurred image spectrum when estimating the fuzzy angle, which affects the quality of image restoration. In addition, when using the traditional Radon transform to detect fuzzy angles, the area of bright spots in the picture is large, which will cause large fluctuations in the detection results. Therefore, this paper proposes a cepstrum analysis method for motion blurred images. Firstly, the gradient cepstrum estimation of the blurred image is given, and the cepstrum is binarized and further refined by setting the gray threshold of the cepstrum. Finally, the line fitting is performed on the refined binary image, and the slope of the fitting line is calculated, so that the point spread function of the blurred image can be accurately and directly estimated, and better image restoration effect can be obtained.