[vtkusers] Homogeneous Transform Matrix to Camera Parameters

[Müller Michael]

Hi all,

i want to convert an estimated camera pose (for an augmented reality application) given as vtkMatrix4x4 to a vtkCamera`s position, focal-point and view-up vector. Position is trivial, but I have some doubts concerning the current calculation of the focal point and the viewUp vector which is as follows:

// camPose: vtkMatrix4x4*
double viewPlaneNormal[3];
double focalPoint[3];
double viewUp[3];

  viewPlaneNormal[0] = camPose->GetElement (2,0);
  viewPlaneNormal[1] = camPose->GetElement (2,1);
  viewPlaneNormal[2] = camPose->GetElement (2,2);

  // focal point
  focalPoint[0] = position[0] - viewPlaneNormal[0];
  focalPoint[1] = position[1] - viewPlaneNormal[1];
  focalPoint[2] = position[2] - viewPlaneNormal[2];

  // ViewUp
  viewUp[0] = camPose->GetElement (1,0);
  viewUp[1] = camPose->GetElement (1,1);
  viewUp[2] = camPose->GetElement (1,2);

The problem is that the camera is slightly at the wrong position when using these calculated parameters. If I manually adjust the camera afterwards with

m_Renderer->GetActiveCamera()->Elevation(2.5);
m_Renderer->GetActiveCamera()->Azimuth(0.5);

the result is the one I want (the real camera and the virtual camera are pretty good aligned).
Actually, the estimated camPose has a small reprojection error, that`s why I suspect I have a mistake in setting up the camera.
The other thing which crossed my mind is the calculation of the view angle which is implemented like this:

  double viewAngle = 2 * atan( ( height / 2 )  / m_FocalPoint[1] ) * 180 / vnl_math::pi;
  m_Renderer->GetActiveCamera()->SetViewAngle(viewAngle);

So the question is if you see any obvious mistakes in these calculations or if you have any ideas why the camera is slightly misaligned.

Thanks in advance,
Michael
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.vtk.org/pipermail/vtkusers/attachments/20100819/aa97f3d6/attachment.htm>