Entire shape scan system with multiple pro-cams using texture information and accurate silhouette creating technique

• Published in: 2015 14th IAPR International Conference on Machine Vision Applications (MVA) pp. 18 - 21
• Main Authors: Oki, Yoshinori, Visentini-Scarzanella, Marco, Wada, Tomohito, Furukawa, Ryo, Sagawa, Ryusuke, Kawasaki, Hiroshi
• Format: Conference Proceeding
• Language: English
• Published: MVA organization 01.05.2015
• Subjects: Accuracy; Cameras; Image color analysis; Image reconstruction; Shape; Three-dimensional displays
• DOI: 10.1109/MVA.2015.7153123

Recently, many approaches have been proposed for complete 3D reconstruction of moving objects. Particularly, one-shot reconstruction using a projector-camera system (pro-cam) has attracted considerable attention as it can reconstruct the full 3D shape from a single image and it is therefore suitable for capturing moving objects. However, one-shot measurement methods have the issue that the projection pattern is easily affected by the texture of the target object and its background. Furthermore, background subtraction can be inaccurate due to the shadow cast by the target object. In this paper, we propose a method to solve these issues using texture estimation and online background synthesis techniques, by relying on a dictionary of pre-captured images. By applying these methods to our multi-view pro-cams, we can demonstrate the robustness of our approach for one-shot reconstruction of moving objects, with reduced sensitivity to texture and cast shadows.