“Depth image based 3D human modeling resolving self-occlusion” by Jang, Cho, Seo, Park and Lee

  • ©

Conference:


Type(s):


Title:

    Depth image based 3D human modeling resolving self-occlusion

Presenter(s)/Author(s):



Abstract:


    The current digital broadcasting system provides a high-quality digital multi-media, but with the rapid advancement of digital technology, the end-users expect to see more realistic contents and even multi-modal interactions with the media. The 3D dynamic objects can play a significant role in these real scenes, so that many researchers have worked on the generation of 3D contents. Among them, we used a depth camera system that can capture both the color and depth data with 60 fps that are compatible to use in a real broadcasting system. However, it suffers from a ’self occlusion’ problem; when it captures a dynamic object a part of the object is occluded by another part of the object. This limits us to model the occluded region, for example, separating an arm from the body. We propose a method to generate a dynamic 3D human model that can be used in 3D broadcasting by resolving the self-occlusion problem.

References:


    1. Domiter, V. 2004. Constrained delaunay triangulation using plane subdivision. In Proceedings of the 8th Central European seminar on computer graphics, Budmerice, 105–110.
    2. Peter, L. 2003. Filling holes in meshes. In Eurographics Symposium on Geometry Processing, 200–206.


ACM Digital Library Publication:



Overview Page: