CellID "stealing" when a cell is missing from a frame in time series datasets

[naikymen]

Description

This happens often when using small images, either because:

• these images have few cells (one to three)
• the cells are missing because they are near the image border
• or simply because they are small in pixels (i.e. 60x60).

We think that CellID might be using the found cell's x/y pos to align the image (instead of the entire image itself).

Example

Time A:

Time B:

Result: note that the yellow line connects points with the same cellID along frames, and while the yellow trace jumps to the other cells position, a new cellID is generated for the remaining mask.