连续子邻域内的鲁棒双边滤波  

Robust Bilateral Filter in Consistent Sub-neighborhoods

作　　者：肖秀春[1,2] 王章野[3] 张雨浓[1] 姜孝华[1] 彭群生[3] 

机构地区：[1]中山大学信息科学与技术学院,广州510275 [2]广东海洋大学信息学院,湛江524025 [3]浙江大学CAD&CG国家重点实验室,杭州310027

出　　处：《计算机科学》2011年第3期275-278,共4页Computer Science

基　　金：国家自然科学基金(60775050);浙江大学CAD/CG国家重点实验室开放课题(A0908)资助

摘　　要：提出一种连续子邻域内的鲁棒双边滤波算法(Robust Bilateral Filtering)。首先,利用自适应区域生长方法在图像局部邻域中分割出种子像素的连续子邻域;然后,在该连续子邻域中采用改进的双边滤波算法对种子像素值进行平滑处理。为了提高算法的鲁棒性能,类似非局域均值滤波算法(Non-Local Means Filtering),以像素空间临近度和像素局部窗口相似度定义该滤波器核函数。算法结合了双边滤波和非局域均值滤波的优点,且在连续子邻域内进行去噪处理相对可获得更为合理的图像效果。仿真实验表明,该算法具有良好的去噪效果,同时较好地保留了图像的细节特征。A robust bilateral filtering algorithm based on consistent sub-neighborhoods was presented.Firstly,employ adaptive region growing method to split local neighborhood of the seed pixel into consistent sub-neighborhoods.Then,within one of the sub-neighborhoods,smooth the value of seed pixel by using improved bilateral filter algorithm.To enhance the robust performance,by following the definition of non-local means filtering,define kernel function of this filter based on geometric closeness and local neighborhood window similarity.Because this algorithm combines both the advantages of bilateral filter and non-local means filter,and also denoises within consistent sub-neighborhoods,it can gain a more reasonable image effect.Simulation experiments demonstrate that the presented algorithm can remove noise effectively,and simultaneously preserve detailed feature of the image.

关 键 词：图像去噪 双边滤波 非局域均值滤波 子邻域 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]