Stabilizing Using CameraTracker

If you have a NukeX license, you can stabilize footage using CameraTracker. CaraVR adds two new options to the standard CameraTracker Export dropdown:

• C_SphericalTransform (stabilise) is useful for stabilizing stitches from other packages, as well as from CaraVR, but it does add another filtering hit during processing which can affect image quality. See C_SphericalTransform Stabilization for more information.

• C_SpericalTransform (Metadata only) can be used for CaraVR stitches to reduce filtering hits, as it edits the metadata passed down from the solver directly. See Transforming Using Metadata for more information.

Note:  You can also handle metadata using C_SphericalTransform. See Transforming Using Metadata for more information.

C_SphericalTransform Stabilization

1.

Render a preview stitch from C_CameraSolver to work on. This step cuts down processing overheads considerably.

2.

Attach a C_SphericalTransform node downstream of the preview stitch.

CameraTracker was developed with standard rectilinear footage in mind, so feeding the latlong preview straight in may cause problems for the tracking process.

3.

Set the Output Projection dropdown to Rectilinear.

Tip:  You can reduce the time taken to track the footage by adjusting the Width control, which cuts down the resolution that CameraTracker has to process. For stabilization purposes, you only need rough rotation data, not pixel-perfect tracking data.

4.

Click the Output tab and make sure that the Focal Length and Sensor controls match those in the C_CameraSolver properties.

5.

Attach a CameraTracker node to the C_SphericalTransform and open its Properties panel.

6.

Track and solve the camera as described in Camera Tracking.

When the process is complete, the Viewer shows the computed tracks and their associated solve error. CameraTracker calculates the solve error for each view in the script, but the main, or Principal View in stereo scripts, is the important view.

7.

Click the Export dropdown and select C_SphericalTransform from the list.

C_SphericalTransform is set up to invert the rotation data extracted from the camera track.

8.

Click Create to connect the new node to the node tree.

The CameraTracker and C_SphericalTransform nodes are linked using an expression, which means that any changes made to the track and solve data in CameraTracker are automatically passed to the C_SphericalTransform node.

9.

The final step is to connect the C_SphericalTransform node to the source latlong preview because the camera data was created in rectilinear space.

10.

Play through the footage to check that the stabilization was successful.