针对由相机与所摄景物之间发生相对位置移动所导致的图像运动模糊, 提出了一种鲁棒的基于单幅运动模糊图像的盲反卷积算法。该方法首先通过预测图像中的较强边缘信息, 实现用简单、易于求解的优化问题在傅里叶域中快速、准确地估计出点扩散函数。然后利用得到的点扩散函数, 使用基于梯度约束的非盲反卷积算法复原清晰图像,同时采用一种新的边缘保持滤波器—导引滤波来消除噪声并抑制振铃效应。实验结果表明: 本文的算法能够快速地从单幅运动模糊图像复原出具有清晰边缘和纹理的高质量图像, 并且运算时间不超过20 s。

To modify the image motion blurring induced by the relative position motion between a camera and a subject, a robust blind deconvolution method based on single image motion blurring was proposed. Firstly, the strong edges of a latent image was predicted. By constraining image gradient, the point spread function in frequency domain was estimated fast and robustly by this method. After that, the non-blind deconvolution algorithm based on the gradient constrain was used to restore the latent image. Meanwhile, a new edge preserving filter-guided filter was used to suppress the ringing and noise. Experiments show that the proposed method can restore a high quality image with clear edges and tissues from single motion blurring image, and the operation time is less than 20 s.