|Title||Image-Based Rendering of Range Data with Estimated Depth Uncertainty
|in||IEEE Computer Graphics & Applications (to appear)|
Christian Hofsetz, Kim Ng, Nelson Max, George Chen, Yang Liu, Peter McGuinness |
|Keyword(s)||Image-Based Rendering, Point-Based Rendering|
|Publisher||IEEE Computer Society|
It is difficult to get accurate depth information with computer visionís 3D estimation techniques. When the estimated depths are uncertain, conventional image-based rendering algorithms will fail and the view rendering quality will suffer. In this paper, we demonstrate a new image-based rendering algorithm which can render good quality views even when the estimated depths are uncertain. Instead of only using the depth information per pixel, we compute what we call a depth uncertainty region around it. We show how to extract the depth from the input images and how to compute the uncertainty region. We show how to render depth with uncertainty by splatting 3-D ellipsoidal Gaussian kernels.