Geometric Inpainting of 3D Structures

Pratyush Sahay, A. N. Rajagopalan; Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2015, pp. 1-7

Abstract


In this paper, we address the problem of inpainting in 3D digital models with large holes. The missing region inference problem is solved with a dictionary learning-based method that harnesses a geometric prior derived from a single self-similar structure and online depth databases. The underlying surface is recovered by adaptively propagating local 3D surface smoothness from around the boundary of the hole by appropriately harvesting the cue provided by the geometric prior. We showcase the relevance of our method in the archaeological domain which warrants 'filling-in' missing information in damaged heritage sites. The performance of our method is demonstrated on holes with different complexities and sizes on synthetic as well as real examples.

Related Material


[pdf]
[bibtex]
@InProceedings{Sahay_2015_CVPR_Workshops,
author = {Sahay, Pratyush and Rajagopalan, A. N.},
title = {Geometric Inpainting of 3D Structures},
booktitle = {Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops},
month = {June},
year = {2015}
}