Fire
-
Results: Ours
Our method using our per-view geometries, our matting, and our frequency-based blending.
-
Results: ULR [Buehler et al. 2001] (videos)
[Buehler et al. 2001] is used to blend the input videos frames on the MVS reconstruction.
-
Results: ULR [Buehler et al. 2001] (looped videos)
[Buehler et al. 2001] is used to blend the input videos frames on the MVS reconstruction. Videos are looped by our method in a preprocess step.
-
Results: Our blending method using the MVS geometry
Our blending method using input videos temporal laplacian pyramids and the MVS geometry.
-
Results: Our geometry and matting with ULR blending
Our method using our per-view geometries, our matting, and the blending from [Buehler et al. 2001].
-
Results: Our method without the depth propagation step
Our method using our per-view geometries but without the depth propagation step.
-
Results: Ours using first depth
Our method using our per-view geometries but taking the first depth along the volume instead of the median.
-
Results: Our background
[Buehler et al. 2001] is used to blend our extracted background videos on the MVS reconstruction.
-
Results: Textured Mesh
Textured mesh from [Reality Capture].
-
Results: Top View
MVS reconstruction with calibrated cameras (blue) and the current synthesized view (in red).
-
Results: ULR [Buehler et al. 2001] (images)
[Buehler et al. 2001] is used to blend the input images on the MVS reconstruction.