Abstract:

The smoothing of cloud points is an important topic in the field of digital geometry processing. Applications based on huge sampled points require fast processing speed and high quality smoothing quality. The traditional CPU based methods deal with every point in a serial manner which leads to great time consumption. a novel approach using graphics processing units(GPUs) is proposed for cloud points smoothing processors in this paper. Many covariance matrices are organized into a large scale spare matrix that is compressed in several textures. The least eigenvalues and corresponding eigenvectors of the matrices are calculated in pixel programs using GPU which has the powerful parallel processing capability. Then the smoothing speed and normal direction can be determined. Experiments are conducted in a PC with the GeForce 8600GTS graphic card. The results show that the efficiency of smoothing processing is improved greatly by using the GPUbased algorithm. Therefore, they well support the applications of the fast processing of cloud points.

Key words: graphics processing unit;smoothing;general computation;neighborhood;covariance matrix