3D reconstruction of cultural heritages: Challenges and advances on precise mesh integration

摘要

Several methods perform the integration of multiple range scans of an object aiming the generation of a reconstructed triangle mesh; however, achieving high fidelity digital reconstructions is still a challenge. That is mostly due to the existence of outliers in the acquired range data, and their harmful effects on the integration algorithms. In this work, we first discuss artifacts usually found on real range data captured with 3D scanners based on laser triangulation. Following that there is the assessment of two widely used volumetric integration techniques (VRIP and Consensus Surface) and our suggested improvements. We also present a novel, hybrid approach that combines strengths from both VRIP and Consensus Surface, named IMAGO Volumetric Integration Algorithm (IVIA). Our novel algorithm adds new ideas while improving the detection and elimination of artifacts. Further, IVIA works in close cooperation with the subsequent hole filling process, which greatly improves the overall quality of the generated 3D models. Our technique leads to better results when assessed in different situations, when compared to VRIP, Consensus Surface, and also to a well known state-of-the-art surface-based method, Poisson Surface Reconstruction.