Robust denoising of unorganized point clouds

A point clouds denoising method based on moving least-squares is presented in this paper. The moving least-squares method has a good ability to denoise a point cloud in 2D. However, there are some serious disculties associated with the moving least-squares technique for some cases, such as varying thickness of the point set and the effects from unwanted neighboring points. Lee improves moving least-squares technique using Euclidean minimum spanning tree, region expansion and refining iteration. In this paper, we take some modification of Lee´s method by using a simple method based on the correlation of the point set to find the adaptive weighting parameter, and extends it to 3D. Experimental results show that our new approach is robust and effective on denoising unorganized point clouds.