Carlos, the 3d estimator estimates the pose between two 3d point clouds, 
usually from stereo devices.

The 2d estimator is a bit of a misnomer.  It matches a 2d image against 
a set of 3d points, using OpenCV's PnP algorithm.  It is what you want 
to use when you already have a set of reference 3d points (and 
associated visual features), and a monocular image.

If you want to match one image to another, to find the essential or 
fundamental matrix, then OpenCV has a function for that.

Cheers --Kurt

On 1/7/2011 7:39 PM, Carlos Jaramillo wrote:
> What are the differences between posest's pe::PoseEstimator2d  and
> pe::PoseEstimator3d? Which one should be used for a monocular camera? I
> believe PoseEstimator2d  is based on the SFM method (Structure from
> Motion). For some reason, I get results only with stereo images and
> PoseEstimator3d. Does any one have any example of posest for a monocular
> camera?
>
> Thank you all!
>
> Carlos
>
>
>
> _______________________________________________
> ros-users mailing list
> ros-users@code.ros.org
> https://code.ros.org/mailman/listinfo/ros-users