Abstract: An algorithm of blind image deconvolution is proposed for defocus blurred images. Firstly, Hough transform is applied to detect the line edges in the defocus image. Then the step-edges or approximate step-edges are located, based on the image spatial statistical characteristic and the modified Grubbs method. The line spread function is calculated by using the detected step-edges, and the radius of the defocus blur is obtained by adopting the relationship between the radius and the line spread function. Finally the defocus image is restored by Wiener filter method. Tested on the real defocus blurred photographs, the experimental results show that the proposed algorithm can detect the step-edges or approximate step-edges accurately, and improve the identification precision of blur radius and the quality of restored images. The algorithm has been applied in the practical detection forensics work successfully.

Key words: defocus blur, step edge, line spread function(LSF), point spread function(PSF), blind image deconvolution

摘要： 针对离焦模糊图像, 提出了一种盲复原算法。该算法首先利用Hough变换检测出离焦图像中的直线边缘, 然后基于图像的空域统计特性和修正的Grubbs检验法, 定位出阶跃或近似阶跃直线边缘, 在此基础上自适应计算出线扩散函数, 最后利用线扩散函数求取离焦模糊半径, 进而用Wiener滤波完成了图像的复原。实验结果表明, 对真实的离焦模糊图像, 该算法能够准确地检测和定位出阶跃或近似阶跃边缘, 提高离焦模糊半径的鉴别精度和图像的复原效果, 已在实际刑侦取证工作中获得较为成功的应用。

关键词: 离焦模糊, 阶跃边缘, 线扩散函数(LSF), 点扩散函数(PSF), 图像盲复原