一种基于区域一致性的平滑滤波算法

文章提出了一种基于区域一致性的平滑滤波方法,旨在滤除受污染图像混合噪声的同时更好地保留图像的边缘信息。该算法将区域样本集按4种不同方向划分为4对样本子集,然后比较每对子集的亮度散布和并选定出一致性最好的子集对,最后由两样本子集均值的离差与各自散布的关系来判断区域是否包含明显的边缘信息,若有则对散布较小的子集局部均值以保护边缘。实验表明,该算法在具有比均值和中值滤波器更好的混合噪声抑制性能的同时较好地保留了图像的边缘信息。此外,该算法运算量适中,具有较好的实时性能。

In this paper, an algorithms for smooth filter based on the region harmony is proposed to retain edges information of an image while reducing the image noise. Firstly, four pair of subset samples are obtained by divided filtering region in four directions. Then, comparing the sum of every pair of subset samples' intensity interspersion, choose the pair of subset samples with the best region harmony. Finally, according to the relationship between the dispersion of the mean and the intensity interspersion of the pair of subset samples, make sure whether there is obvious edge information in the region. If so, the mean of the one with smaller intensity interspersion is the filtering result. The experiment results show that it has a better filtering performance than mean filters and median filters while protecting the image edges. Besides, it has moderate computation and preferable real-time capability.