基于正则化方法的图像盲去模糊

针对标准化稀疏先验的正则化方法估计复杂模糊核时的不准确性, 引入图像的预处理, 提出了一种图像盲去模糊的新方法。该方法将图像盲去模糊分为三个步骤：利用双边滤波器和冲击滤波器对图像进行预处理, 使得图像的噪声降低、边缘突出, 有利于模糊核的估计; 对预处理后的图像, 利用基于标准化稀疏先验的正则化方法估计模糊核; 根据估计出的模糊核利用TV正则化方法对图像进行非盲去卷积。采用快速迭代收缩阈值算法和快速总变分图像复原算法分别求解模糊核估计模型和图像非盲去卷积模型。实验结果表明, 针对单幅模糊图像, 该方法可以估计出准确的模糊核, 对噪声具有鲁棒性, 并且提高了图像复原速度, 具有较好的图像恢复效果。

Blind image deblurring based on regularization method

Aiming at the inaccuracy of estimating the complex blur kernel, which was based on the regularization method of the image normalized sparse prior, this paper introduced the image preprocessing and proposed a novel method of blind image deblurring. The method divided blind image deblurring into three procedures. Firstly, it adopted the bilateral filter and impact filter to preprocess the image. It could strengthen image edge and reduce noise, which would be good for estimating blur kernel. Then, estimated blur kernel based on the regularization method of the image normalized sparse prior. Finally, it adopted TV regularization method in non-blind deconvolution processing. Fast iterative shrinkage-thresholding algorithm solved the model of image blur kernel estimation, simultaneously fast total variation image restoration algorithm solved non-blind deconvolution mode. Experimental results show that the proposed method is fast and very robust, and can accurately estimate the blur kernel. Moreover, it can improve the restoration effect of the blur image.