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

In many applications, denoising is necessary since point-sampled models obtained by laser scanners with insufficient precision. An algorithm for point-sampled surface is presented, which combines fuzzy c-means clustering with mean shift filtering algorithm. By using fuzzy c-means clustering, the largescale 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.