Recent work in my research group has resulted in three processes that
might be of interest to other researchers in computer vision and
graphics.The theme that links the processes is 3D data and scene
understanding or analysis.
The three themes are:
1) When building virtual 3D models of scenes, complete data acquisition can
be practically impossible, but inference of much occluded data is
possible.We present a new technique where the scene is its own
source of information for 3D reconstruction.
2) When the color texture mapping is acquired for a 3D scene, the colors can
be very inconsistent due to lighting, shadows, interreflections, etc.
We present a new technique for computing more consistent coloring
across the texture maps.
3) The addition of colour information to the computation of 3D range/scene
flow is proposed to improve its accuracy and robustness to ambiguities. This
is applied in the form of additional optical flow constraints from aligned
colour image data. Combining constraints gives improved velocity
displacement fields over using depth alone, or in using depth plus intensity.