基于非均匀模糊核的RL改进算法

针对传统均匀模糊模型下的Richardson Lucy（RL）算法存在振铃效应和细节丢失等问题，提出了一种基于非均匀模糊模型下的改进RL算法。基于Yu-Wing的射影运动模糊模型，通过在迭代过程中采用基于局部极值分解的方法提取含有图像结构但无振铃的边缘图像，将模糊图像和上述提取的边缘图像作为输入图像，用添加了局部先验项的RL法对输入图像进行去模糊。实验验证了算法抑制振铃效应的有效性，同时很好地保留了图像细节，表明非均匀模糊模型在相机抖动产生的模糊图像去模糊中更为有效。

Improved RL method based on non-uniform kernel

In view of the problem of ringing effect and missing details in Richardson Lucy（RL） algorithm under conventionally uniform blur model, an improved RL algorithm based on the non-uniform blur model is put forward. Under the blur model of projective motion path based on Yu-Wing, edge image which has structure and no ringings is extracted, using the method based on local extremum decomposition in the iterative process. Making the blurred image and the edge image as the inputs, then the inputs using the RL method with partial prior terms are deblurred. Experimental results demonstrate the effectiveness of the algo- rithm to suppress the ringing effect, while preserving the image details, and indicate that the non-uniform blur model is more effective to deblurring in camera shake.