This forum is disabled, please visit https://forum.opencv.org
 | 1 | initial version |
During the transformation from 3D to 2D you are losing the depth information. All you can do with the matrices that you have, is to transform a 2D pixel into a 3D line where every point on this line would be projected onto the same 2D pixel.
You will definitely need additional information to reconstruct a 3D point. For example an additional view of the same scene at the same time (stereo vision).
| | 2 | No.2 Revision |
During the transformation from 3D to 2D you are losing the depth information. All you can do with the matrices that you have, is to transform a 2D pixel into a 3D line where every point on this line would be projected onto the same 2D pixel.
You will definitely need additional information to reconstruct a 3D point. For example an additional view of the same scene at the same time (stereo vision).
EDIT
The homography transforms one plane in 3D space into another plane in 3D space. You cannot get absolute coordinates from this, because the homography is not unique.
You can get some information from a single view, if you have some additional information about the view (if you have a calibration object in the view, or if you can use structured lighting, etc.)
I recommend searching for "single view geometry" and if you want to read an excellent book on this topic, I recommend "Multiple View Geometry in Computer Vision" by Hartley & Zisserman, which also has a chapter about single view geometry.
