Kun Liu

Sphere Packing Aided Surface Reconstruction for Multi-View Data

Kun Liu1,2 Patricio A. Galindo1 Rhaleb Zayer1

1INRIA, France

2University College London, United Kingdom

10th International Symposium on Visual Computing (ISVC), 2014


Abstract:

Surface reconstruction has long been targeted at scan data. With the rise of multi-view acquisition, existing surface reconstruction techniques often turn out to be ill adapted to the highly irregular sampling and multilayered aspect of such data. In this paper, a novel surface reconstruction technique is developed to address these new challenges by means of an advancing front guided by a sphere packing methodology. The method is fairly simple and can efficiently triangulate point clouds into high quality meshes. The substantiated experimental results demonstrate the robustness and the generality of the proposed method.

Paper:

PDF

Images:

algorithm

Figure 1: An illustration of the algorithm.

operations

Figure 2: Three operations applied for advancing the current front (yellow): (a) ear cutting; (b) point addition; (c) merging fronts.

res_church

Figure 3: A point clouds from [25] is triangulated using our proposed method. (a) displays the resulting mesh. (b) is a close-up view.

pr_vs_ours

Figure 4: A comparison of Poisson reconstruction (a) and our proposed method (b) is illustrated. The input point cloud is same to the one used in in Figure 3a and the two meshes are displayed in the same close-up view. (c) and (d) are two corresponding histograms about triangle angle values.

evaluation

Figure 5: The reconstruction results of BP, PR and YO in [18,21,2] respectively, as well as our methods are evaluated using the method proposed in [25]. The images show the variance weighted depth difference. Red pixels represent errors larger than 30σ. Green pixels represent the missing scan data of the ground truth. The relative errors between 0 and 30σ are displayed using gray scale from 255 to 0.