I need some clarification on the non-linear weighting m, M and P in APS processing panel (see the attached image). Specifically, how they are implemented in the APS estimation and what is their mathematical meaning?
we did not publish anything on this topic. However, the concept is simple: is e.g. 0.7 coherence high or low? it depends! It depends on how many images are in your dataset, on the baseline distributions, it depends on the processing options you have chosen. So, firstly Sarproz runs some simulations to find reference numbers (m the minimum below which you can consider your point as totally noisy, M the maximum over which you can consider your point as totally reliable and p, a value in between), then those numbers are used to stretch the estimated coherence towards zeros (noise) and ones (reliable points). The obtained “stretched” values are then used to drive the inversion (or integration) of the spatial graph (PSC connections).
These topics are treated in the Sarproz course.