< Previous | Contents | Next >
Clips That Don’t Work Well for Camera Tracking
Even though the Camera Tracker is somewhat automatic, it sometimes needs your help. If you can identify potential issues before you even track or solve the shot, you can save yourself much time. Certain types of clips will cause more significant problems for camera tracking than others. Some are fixable, while for others you just have to admit defeat and figure out another solution. Here is a list of the types of shots to look out for, as they can be big headaches for camera tracking:
— Lack of depth: Camera tracking requires parallax in a clip in order to work. You must be able to identify objects further away and objects that are nearer as the camera moves. If everything is at the same distance from the camera, there is no way to calculate depth. In this case, it’s better to skip the Camera Tracker node and find another solution.
— Locked-off shots: If the camera does not move, there is no way to calculate which objects are closer and which are nearer. Again, don’t spend too much time in this situation; it is better to skip the Camera Tracker node and find another solution.
— Tripod pans: Similar to a locked-off shot, there is no way to calculate which objects are closer and which are nearer from a pan that remains centered on a locked-off tripod. Skip the Camera Tracker node and find another solution.
— No detail: Clips like green screens without tracking markers lack enough detail to track. If you are lucky enough to be involved in the shooting of these types of shots, including tracker markers
makes it much easier to get a good track. Without detail, camera tracking will fail and you will need to find a more manual solution.
— Motion blur: Fast camera motion or slow shutter speeds can introduce motion blur, which will make it difficult to find patterns to track. It’s worth trying shots like these to see if there are enough details to get a good solve, but know when give up and turn to another solution.
— Rolling shutter: CMOS-based cameras sometimes introduce distortion due to the shutter capturing different lines at slightly different times. This distortion can create significant problems for camera tracking. Sometimes it is possible to create motion vectors with the Optical Flow node to create new in-between frames without the wobble distortion of the rolling shutter. Then you can use the corrected image to connect to the Camera Tracker.
— Parallax issues: When objects at different distances in a shot overlap in the frame, the overlapping area can be misinterpreted as a corner. Having a tracker assigned to an overlapping angle like this will cause errors as the parallax starts to shift and the overlapping area slides. This can be solved in Fusion by removing that tracker before running the solver.
TIP: Some shots that cannot be tracked using Fusion’s Camera Tracker can be performed in dedicated 3D camera-tracking software like 3D Equalizer and PF Track. Camera tracking data from these applications can then be imported in the Camera3D node in Fusion.
TIP: Some shots that cannot be tracked using Fusion’s Camera Tracker can be performed in dedicated 3D camera-tracking software like 3D Equalizer and PF Track. Camera tracking data from these applications can then be imported in the Camera3D node in Fusion.
TIP: Some shots that cannot be tracked using Fusion’s Camera Tracker can be performed in dedicated 3D camera-tracking software like 3D Equalizer and PF Track. Camera tracking data from these applications can then be imported in the Camera3D node in Fusion.
— Moving objects: It’s difficult to capture a shot where objects in the clip do not move. People, cars, animals, or other object may move in and out of a shot. These objects move independent of the camera movement and must be eliminated or they will cause solving errors. You can fix these issues by masking out objects that are not “nailed to the set.” The masks are then connected to the Track Mask input on the Camera Tracker node.