Link overlapping objects in binary images

I have binary images of the same size in which I would like to track the different objects and see whether they have grown, shrunk, merged or dissapeared.

Due to the way bwlabel works I would get the following result and would like to find an efficient way to see which object in image 1 is linked to which object in image 2. Since the objects do not move much I should be able to work with some algorithm to check if objects overlap but for 1000 images with around 1000 objects each I feel as if I need a smart way to not waste too much computational power and time.

Therefore I would be happy about any input what a smart way would be to tackle that problem. If the resulting array would be like this, it would be amazing.

Result = [1 2

2 1

3 3

4 0

5 3

0 4]

[DGM]

Have you considered segmenting the image as a volume? As a volumetric image, the connectivity between blobs in adjacent slices becomes apparent. Instead of having to get correspondent label ordering on many different frames, you can just label the volumetric image and all slices of an object will have the same label. The drawback is that it may be difficult to deal with the memory requirements of a MxNx1000 image. That might make this suggestion problematic.