CODEWITHSUNDEEP
quaternionscamera-matrix
KBowser
KBowser

Reputation: 41

Camera extrinsic parameters as dual quaternion

Does it make sense to represent the pinhole camera extrinsic parameters as a dual quaternion instead of the classical 4x3 matrix when projecting a 3D world point to the image plane of the camera? I also need to do the inverse (from 2D + depth to 3D world) so this is why I thought this representation could be useful to save some time, as the inverse of a dual quaternion it's not too difficult to calculate.

Upvotes: 0

Views: 562

Answers (1)

Martin Prazak
Martin Prazak

Reputation: 1605

Be aware that dual quaternions can't represent non-rigid transformations well (the "rotational" part is assumed to be a unit quaternion, the translational part is essentially a translation vector - see here).

This means that for extrinsic parameters, dual quaternions are a good representation (albeit a bit unorthodox one). However, from your question it seems that you want to represent also the intrinsic parameters in the same way (2D+depth to 3D and back). Intrinsic parameters are a non-rigid transformation (they contain scale, and sometimes skew), and dual quaternions cannot be easily used to represent them.

Upvotes: 2

Related Questions