I am not sure if I think about the same thing, but since we use multiple images for PSI, a single coherence map would be a bit misleading, since it would only communicate two of many coherence values of the time series and image pairs…and the selection of only one image pair would be very subjective.
What you could think of is to make many many coherence maps, stack them and then reduce them maybe by selecting the min to make a conservative approximation of the coherence over time. However, that result would probably lead to a very simple PSI analysis and the coherence map would be very similar to the PSI measurement points, thus I am not sure if we mean the same. Can you provide an example in some publication you have read or explain to me what you mean and where I might be wrong?