基于细节保存的图像深度脉冲噪声去除算法

Terrible Impulse Noise Removal Based on Adaptive Median Filter and Detail-Preserving Regularization

提出了一种融合自适应中值滤波与细节保存机制的图像深度脉冲噪声去除方法。该方法主要采用两个步骤,首先使用自适应中值滤波来最大可能地确定图像中的候选噪声点,然后根据特定的针对候选噪声点的正则化优化方法来复原被污染的图像。实验结果表明,与传统的自适应中值滤波相比,本文方法具有明显优越性,可以完成被脉冲噪声污染程度达90%的图像噪声去除工作,具有一定的实用意义。

This paper proposes an algorithm based on adaptive median filter and detail-preserving regularizationIn the first phase,an adaptive median filter is used to identify pixels which are likely to be noise candidates.In the second phase,the image is restored using a specialized regularization method that applies only to those selected noise candidates.The experiment results demonstrate that this method has the superiority of reliability compared with the traditional adaptive median filter,even so that the noise level is as high as 90%.