中文摘要：

为了在有效滤除椒盐噪声的同时更好地保持图像细节，提出一种基于滤波值二次修正的图像椒盐噪声滤波算法。该算法首先利用基于阈值的自适应中值滤波改进算法实现对噪声点的检测与初级滤波，然后对滤波值采用基于方向信息的多阈值自适应滤波算法，实现对初级滤波值的二次修正。算法结合了基于阈值的自适应中值改进算法良好地噪声检测性能及基于方向信息的多阈值自适应滤波算法的细节保护能力。实验结果表明，在客观评价指标和主观视觉效果方面，新算法均取得了优于传统自适应中值滤波算法及其改进算法的效果。

英文摘要：

To preserve the detail while removing salt-and-pepper noise effectively, an algorithm based on second correc- tion of filter value is proposed. Improved algorithm of adaptive median filter based on threshold value is used to detect the noise and filter, and then adaptive median filter algorithm with multi-threshold based on directional information is used to correct the filter value. The algorithm takes the advantages of noise detection performance of improved adaptive median filter algorithm and detail preserving performance of multi-threshold adaptive filter algorithm based on directional infor- mation. The experimental results show that the new algorithm is superior to traditional adaptive median filter algorithm and improved algorithm in terms of both objective evaluation and subjective visual effect.