Point Cloud De-Noising Based on Three-Dimensional Projection
Rui Li, Qi Yan, Liping Liu · 2015
Through K-nearest neighbor algorithm based on k-d tree, the noises in the disordered and three-dimensional point cloud are removed. The value of the k in the K-nearest neighbor algorithm is discussed in detail. By three-dimensional projection, the dimension of k-d tree is reduced. So efficiency of de-noising is improved. The original depth image corresponding with the point cloud data is converted to gray scale images. After binarization and extracting foreground region, the correspondence between the foreground region and the point cloud data helps to remove the noise from specified perspective. By a best perspective projection, three-dimensional point cloud is simplify to two-dimensional point cloud. Using the K-nearest neighbor algorithm to remove the rest of noises. Consequent of de-noising is good.