Probabilistic merging of sources

[MatthewWhiting]

From a discussion with Russell:

It may be possible to assign a probability to the merger of two objects. One would do this by looking at the voxels in between objects (NB - how is "in between" defined?), and looking at the probability of them each being just due to noise.

If you can reject the hypothesis that there is just noise between them at some level, then you can merge them. (Q: how do you assign the intervening voxels to the object? Only those above some threshold? What is that threshold?)

This raises the possibility of providing a nested list of detections, where both the original, separated detection list *and* the final merged list is provided, where the merged list has pairwise probabilities of objects being merged. The fact that the probs have to be pairwise may be complicated.