Shape Segmentation by Approximate Convexity Analysis

(1 view)	(2 views)	(3 views)	(5 views)

Figure: Results of our segmentation algorithm applied to a progression of incomplete point clouds with an increasing number of scanned views. With only two views, the parts are already meaningful, despite the significant amount of missing data.

Abstract

We present a shape segmentation method for complete and incomplete shapes. The key idea is to directly optimize the decomposition based on a characterization of the expected geometry of a part in a shape. Rather than setting the number of parts in advance, we search for the smallest number of parts that admit the geometric characterization of the parts. The segmentation is based on an intermediate-level analysis, where first the shape is decomposed into approximate convex components, which are then merged into consistent parts based on a non-local geometric signature. Our method is designed to handle incomplete shapes, represented by point clouds. We show segmentation results on shapes acquired by a range scanner, and an analysis of the robustness of our method to missing regions. Moreover, our method yields results that are comparable to state-of-the-art techniques evaluated on complete shapes.

Results

Figure: Segmentations obtained with our non-parametric algorithm on a collection of point clouds acquired with a range scanner.

Figure: Segmentations obtained with our non-parametric algorithm on a collection of triangle meshes. A quantitative evaluation on the full segmentation benchmark with 380 meshes is reported in the paper.

Figure: Convergence of the segmentation results as the point clouds are complemented with an increasing number of scanned views. Notice how with a small number of views, the results are already of similar quality to the results with several views (in the last column).

Resources

Paper (PDF, 7MB)
MATLAB Code
Datasets (point clouds and incremental views in obj and ply formats)
Results (segmentation results in off and ply formats)

BibTex Reference

@article{vankaick14convseg,
    author = {Oliver van Kaick and Noa Fish and Yanir Kleiman and Shmuel Asafi and Daniel Cohen-Or},
    title = {Shape Segmentation by Approximate Convexity Analysis},
    journal = {ACM Trans. on Graphics},
    volume = {to appear},
    year = 2014,
}
        

Acknowledgements

We thank Hui Huang for providing some of the point cloud models and Chen et al. for the segmentation benchmark. This work is supported in part by the Israeli Science Foundation (grant no. 1790/12) and the U.S.-Israel Binational Science Foundation (grant no. 2012376). Oliver van Kaick is grateful to the Azrieli Foundation for the award of an Azrieli Fellowship.

Last update to the page: April 21, 2014.
Last update to the code: November 23, 2013.