Point Cloud Noise and Outlier Removal for Image-Based 3D Reconstruction

publication
IEEE International Conference on 3D Vision
authors
Katja Wolff, Changil Kim, Henning Zimmer, Christopher Schroers, Mario Botsch, Olga Sorkine-Hornung, Alexander Sorkine-Hornung

Point Cloud Noise and Outlier Removal for Image-Based 3D Reconstruction

From a set of images of a scene(left), multi-view stereo methods can reconstruct a dense 3D point cloud (middle left), which however often suffers from noise and outliers. We propose a simple and efficient filtering method that produces clean point clouds (middle right) that allow for a favorable surface reconstruction (right).

abstract

Point sets generated by image-based 3D reconstruction techniques are often much noisier than those obtained using active techniques like laser scanning. Therefore, they pose greater challenges to the subsequent surface reconstruction (meshing) stage. We present a simple and effective method for removing noise and outliers from such point sets. Our algorithm uses the input images and corresponding depth maps to remove pixels which are geometrically or photo-metrically inconsistent with the colored surface implied by the input. This allows standard surface reconstruction methods (such as Poisson surface reconstruction) to perform less smoothing and thus achieve higher quality surfaces with more features. Our algorithm is efficient, easy to implement, and robust to varying amounts of noise. We demonstrate the benefits of our algorithm in combination with a variety of state-of-the-art depth and surface reconstruction methods.

downloads

accompanying video

acknowledgments

This work was supported by the NCCR Digital Fabrication, funded by the Swiss National Science Foundation, NCCR Digital Fabrication Agreement #51NF40-141853. Mario Botsch was supported by the Cluster of Excellence Cognitive Interaction Technology CITEC (EXC 277) at Bielefeld University, which is funded by the German Research Foundation (DFG).