Fuzzy c-Means and Mean Shift Algorithm for 3DPoint Clouds Denoising

Abstract

In  many  applications,  denoising  is  necessary  since  point-sampled  models  obtained  by  laser  scanners  with  insufficient  precision.  An  algorithm  for  pointsampled surface is presented, which combines fuzzy c-means clustering with mean shift filtering algorithm. By using fuzzy c-means clustering, the large-scale noise is deleted  and  a  part  of  small-scale  noise  also  is  smooth.  The  cluster  centers  are regarded  as  the  new  points.  After  acquiring  new  point  sets  being  less  noisy,  the remains noise  is smooth by mean shift  method. Experimental results demonstrate that the algorithm can produce a more accurate point-sample model efficiently while having better feature preservation.