Tracking segments

The object motion can be too complex to be tracked accurately enough over the whole sequence with the same settings. For this reason, PatchMaker allows the user to split the sequence into several segments where motion is tracked independently:

Each segment, therefore, corresponds to separate tracking procedure determined by:

  • position of the key frame [2],

  • time span (backward [1] and forward [3] tracking bounds),

  • object mask for the key frame, and

  • tracking parameters.

The position of the segment’s key frame is set when the segment is created and cannot be modified later. If you wish to shift the key frame to another frame, you have to create a new segment

 

Data
layer

Each segment has an associated data layer. For the tracked intervals [B, C] indicated by orange color, it consists of interframe motion computed by the tracking algorithm using the parameter values associated with the segment. In intervals that lack computed data [A, D], the motion is simply extrapolated: for any two consecutive frames it is taken to be the same as between the last two frames computed in the same direction (or zero, if no computations was done in this direction).

The segment motion data are uniquely determined by the key frame mask and tracking settings and are discarded when the mask or parameter values are changed. However, changing the tracking bounds does not invalidate the data. (When tracking is rerun over the same segment with adjusted bounds, motion is computed only over those intervals that lack previously computed data.)

 

Segment
composition

See also:

Segment
area

The motion data computed for different segments are combined into the output transformation layer according to how they are vertically positioned in the segment area of the Trackline panel. The data from upper segments override those of lower ones.

Suppose, there are three segments [A] in the segment area as shown above. Then the output transformation layer [B] will combine the segment motion data in a pattern indicated by digits. The output layer is not necessarily continuous and may have gaps consisting of frames not covered by any segment.