Abstract: To improve the resolution of defocused blurred images, the paper presents a blind restoration algorithm of focus blur based on spectrum preprocessing and improved Hough transform. Firstly, the spectrum preprocessing strategy is improved to deal with the blur image spectrum map, and the impacts of noise on the detection of zero point dark circle is reduced. Then the algorithm of Hough transform circle detection is improved to reduce its complexity, meanwhile the accuracy of the estimation of the blur radius is enhanced. Finally, the authors take advantage of the mixed characteristic regularized restoration image model to restore the blur image iteratively, which makes the edge details of the restored image more focused. The experimental results show that the proposed method of blur radius estimation has less average error than other methods. The improved spectrum preprocessing strategy is more advantageous to the detection of zero dark circles. The improved algorithm of the Hough transformation circle detection algorithm has higher accuracy of the blur radius estimation. The proposed algorithm has a better recovery effect on defocused images taken by a small UAV out of focus on the camera. As for the blind restoration of focus blur images, the theoretical analysis and experiment validate the good robustness of the improved blur radius estimation method in this paper.

Key words: blind image restoration, focus blur, spectrum preprocessing, Hough transform, regularization of mixed characteristics