一种改进的Richardson-Lucy正则化图像复原算法

An Algorithm of Modified Richardson-Lucy Regularization Approach for Image Restoration

运动模糊图像复原是模糊图像复原领域中的重要课题。首先,利用一种新的正则化方法有效估计了非参数运动模糊核。然后,在贝叶斯框架下,基于图像像素空间和梯度空间的统计特性,结合自然图像的梯度约束,提出了图像复原的改进RL正则化算法,有效抑制了复原图像中存在的振铃效应。实验结果表明,RL正则化算法在模糊核的估计存在误差时,依然能够得到较好的图像复原效果。

Motion blurred image restoration is the important subject of blurred image restoration, a novel regularization approach is introduced to estimate the non-parameter motion blur firstly in this paper. Then, based on the statistical properties of pixel space and gradient space, a modified Richardson-Lucy （RL） regularization approach for image restoration is introduced, also using the gradient constraints of the natural images. The proposed approach can suppress the ringing effect effectively in the restored image. Then experimental results show that the proposed RL regularization approach can restore the blurry image effectively, even the motion blur is contaminated by measurement errors.