Logic to face vtkCamera at an image should be refactored, and generalised to arbitrary image planes.
In vtk_overlay_widget there is logic in __update_video_image_camera. This positions the camera to face an image, but assumes the x and y axis are fixed.
For applications where you want a vtkActor to move in one window (eg. tracked US probe), but in another window for the camera to always face the image, then the logic in __update_video_image_camera should be generalised to accept images that are at an arbitrary dimension.
Implemented in NifTK here.