一种权重的高阶扩散图像盲复原方法

为了实现对模糊噪声图像的清晰化盲复原,提出了一种权重的高阶扩散图像盲复原方法。首先,在模糊核的估计阶段,运用一种各向异性扩散的图像结构提取策略和shock滤波器将图像中的强边缘准确地提取出来,并利用提取的图像强边缘实现对模糊核的准确估计;然后,在图像的复原阶段,利用高阶扩散模型和权重平衡参数,针对复原图像,提出了一种权重的高阶正则化约束来实现图像的清晰化复原;最后,运用了一种分裂的布雷格曼(split Bregman,SB)最优化迭代策略对提出的方法进行最优化求解。实验结果表明,较近几年的一些具有代表性的图像盲复原方法相比,不仅主观的视觉效果得到了较为明显的改进,而且客观的峰值信噪比也增加了1.0~2.7 dB。

A Weighted High-order Diffusion Method for Image Blind Restoration

In order to recover the blurred-noisy image blindly, a weighted high-order diffusion method for image blind restoration is proposed. First, for the blur kernel estimation, we extract strong edges from the estimated image by using the anisotropic diffusion and shock filter. Then, the extracted strong edges are used for blur kernel estimation. Second, for the image restoration, combining the high-order diffusion model and the weighted parameter, a weighted high-order regularization constraint is proposed for image restoration. Finally, a Split Bregman (SB) optimization strategy is employed to restore the image and simultaneously exactly estimate the blur kernel. Experimental results indicate that the proposed method outperforms some representative image blind restoration methods, not only the subjective vision has the betterment obviously, but also the peak signal to noise ratio improves between 1.0 dB and 2.7 dB.