Graphics Reference
In-Depth Information
Figure 1.13
Some 3D frames computed by the temporal 3D super resolution approach proposed by
Ouji et al. (2011)
towards plausible reconstruction
H
. The 3D model
M
t
cannot be represented by only one 2D
disparity image since the points situated on the fringe change-over have sub-pixel precision.
Also, pixels participate separately in the 3D model since the 3D coordinates of each pixel is
retrieved using only its phase information. Thus, for each camera three 2D maps are created,
defined by the
x
-,
y
- and
z
-coordinates of the 3D points. The optimization algorithm and the
deblurring are applied to compute high resolution images of
x
,
y
, and
z
and texture from the low
resolution images. The final high resolution 3D point cloud is retrieved by merging obtained
3D models that are already registered since all of them contain the 3D sparse point cloud. The
final result is illustrated in Figure 1.13.
1.4.4 Spacetime Faces
The vast majority of stereo research has focused on the problem of establishing spatial corre-
spondences between pixels in a single pair of images for a static moment in time. The works
presented in Davis et al. (2003) and Zhang et al. (2003), which presented nearly identical
ideas, proposed to introduce the temporal axis (available since they process video sequences)
to improve the stereo matching problem. They proposed
spacetime
stereo matching algo-
rithms based on similar ideas. The algorithm proposed in Davis et al. (2003) was tested on
