Particular Bundle Adjustment problem

[GoogleCodeExporter]

(this is more an idea/suggestion, I am not sure if this is the right place to 
post it)

I've integrated Ceres in my 3D reconstruction pipeline in order to solve the 
BAL (bundle adjustment) problem. It works as expected and I am happy with it. 
However, for this particular problem, there is a very important constraint that 
can be applied during the bundle optimization process, which can significantly 
increase the precision of the solution, but I am unable to find a way to do it 
in Ceres. This constraint should be set on the camera intrinsics to keep them 
equal for views taken with the exactly the same camera and settings. Most of 
the time when reconstructing a scene from multiple unordered images, many 
images are taken with the same camera and same setting, so that is expected 
their found intrinsics to be also the same. From my tests implementing this in 
an modified version of SBA greatly improves the robustness of the BA process.

Any suggestions how to implement this in Ceres?

Original issue reported on code.google.com by cdc.seac...@gmail.com on 25 Oct 2012 at 11:30

[GoogleCodeExporter]

The simplest way to do this, would be to create a different parameter block for 
the camera intrinsics, and then use it in all the residuals corresponding to 
images taken with the same camera. 

This is already possible with ceres, as it allows for arbitrary sharing of 
parameter blocks across residuals and handles the corresponding sparsity with 
no problems.

Finally, these kinds of questions are better posted on the mailinglist, the 
issue tracker is for bugs, so I am going to go ahead and close this.

Original comment by sandwich...@gmail.com on 4 Nov 2012 at 2:18

• Changed state: Invalid