基于法向修正的双边滤波点云去噪处理

Point Cloud Denoising Based on Normal Modification and Bilateral Filtering Processing

为了在降噪的同时保持点云模型的几何特征,并较好地处理离群点,针对不同种类的噪音问题,提出了一种基于法向修正的双边滤波点云综合去噪算法。该算法先对点云模型进行空间单元格划分,构造基于单元格最佳连通域,保留内部点云数据并去除点云模型离群点,然后根据双边滤波需要求法线的特点,对点云模型进行法向修正后再进行滤波处理。实验结果表明该算法简单快捷,能够很好地去除离群点,在去噪的同时可以很好地保留点云模型的几何特征,在处理效率和稳定性方面均优于单一的双边滤波算法。

To preserve the geometric features of the point clouds model and get a better result of the outliers in the process,a denoising algorithm for bilateral filtered point clouds was proposed based on normal modification. Firstly,the point clouds model was subdivided into several cells and the optimal interconnecting domains were built based on those cells,thus the point clouds data was preserved and outliers were removed. Secondly,normal features were obtained according to the requirement of bilateral filtering. The normal of the point clouds model was modified and the filtering process was carried out. Experimental results showed that the algorithm was quick,brief,and able to well preserve at removing outliers. It also had a good performance on distinguishing noise and was better than bilateral filtering methods in effectiveness and stability.