Tracking elements to a plate
Tracking elements to a plate can be useful in many ways; its main function is to synchronise the movement of the object in a scene. There are many tracking software’s available, but they all produce the same outcome. A user will select a region within a tracking software of choice, this region is broken up in two, one which contains the pixels that the software will search for and try to match up with, the second region is the search area that the first region is limited to. Without this constraint, the tracking process will deem to be a time-consuming process.
A good tracking software can often calculate sub-pixel offsets. This process is useful in reducing the error. The smaller the sub-pixel match, the more accurate the data, therefore it is required to obtain a good result.
Stabilising a plate
Stabilising a plate uses the same technique as mentioned above, but processes the tracking information to result in a stable plate. It is often used in cases where a locked-off shot is interrupted in some way, the camera operator being bumped for example.
The user will begin by selecting a feature to track to determine its motion; the feature would have to be something which is supposed to be stable within the scene. The information gathered from the tracker would be applied to the plate resulting in stable footage. Noise with the footage can be problematic however, this may cause a jitter when stabilising the shot. There are ways to work around this using the tracks curvature, more explained below.
Choosing a feature to track
Experienced trackers gain an instinct on which pixels to track. Firstly, the user would review the entire footage in search of a feature that is visible throughout the sequence. Secondly, the user would ensure that the feature is consistent within its shape and has distinct edges (a shape that morphs into another would be nearly impossible to track in one go). The colour values of a tracker can also be of equal importance. The tracker’s algorithm will search for pixels that resemble the original sample.
It’s a common misconception that you have to use a single tracking point throughout a sequence, this isn’t correct. If the original tracking point goes out of frame mid sequence, you keep the data you’ve already acquired. Track a new point, and merge the data together (It’s important to note that the second point must still be relative to the first).
Certain scenarios may require the user to temporarily alter elements within footage; this could be altering the luminance values, or soloing colour layers. Doing so may result in having a more accurate feature to track. You can reverse these once you’ve extracted the information you need.
Manual tracking curve
As previously mentioned, the tracking software does most of the work. This being said, in almost all cases the user is required to do additional alterations to the tracking data. The most effective way to check the track is by attaching an object to the tracked point and view the overall result. Another way is to view the curve editor within the tracking software.
The noise can affect auto-tracking results. In the image above you can see that ‘colour-correction, de-grain, and slight blur’ have already been applied. However the noise is still causing the track to be slightly jumpy (see the graph), in this case, it is required to manually examine the curve. The tracking software may have built-in tools which can smooth the curve, if not available, or it doesn’t produce the desired result, then it may be required to create a second curve manually or alter the original curve to produce accurate data.