混合高阶全变差正则化的运动模糊图像盲复原

针对运动模糊图像的盲复原，提出一种基于混合高阶全变差正则化的盲复原方法。该方法首先采用shock滤波器从模糊图像中预测出清晰的图像边缘，并用多尺度策略实现对模糊核由粗到细的准确估计。然后根据自然图像边缘的稀疏特性，将全变差模型的保护边缘特性结合高阶全变差克服平滑区域阶梯效应的优势，对图像进行正则化约束，提出新的混合高阶全变差正则化模型。最后，利用分裂布雷格曼迭代策略对提出模型进行最优化求解。实验结果表明，提出的方法能够很好地保护图像边缘细节，同时有效地抑制平滑区域内振铃和阶梯效应的产生，获得高质量的复原图像。与近几年图像盲复原算法相比，不仅改进了复原图像的主观视觉效果，而且客观上提高了峰值信噪比。

Hybrid high-order total variation regularization method for motion blur image blind restoration

This paper proposed a method based on hybrid high-order total variation regularization method for motion blur image blind restoration. It used shock filter to predict the sharp edges of images and a multi-scale strategy to estimate the blur kernel from a coarse estimation to a fine one. Based on the sparse edges in a natural image, it put forward a new model which combined edges protection features of total variation with the advantage in overcoming the smooth area staircase of high order total variation. In order to solve the model, it used a multi-variable split bregman optimized iterative scheme. The results of experiments show that the proposed method is able to preserve the image edges while avoiding staircase effects and false edges in the smooth regions, and can get high quality images. Compared with recent image blind restoration methods, the proposed method not only improved visual effects but also improved peak signal to noise ratio(PSNR).