I can load a SLC Sentinel product, run the debursting and merge operations in SNAP but it looks like the output of these algorithms are kept in the image space and the complex data is lost. Is it possible to keep the data as complex and if not, is there a description of these algorithms somewhere that I might be able to write my own version that will do so?
What difference does is make when you do deburst & merge? In other words, does it really matter considering that complex data is not well suited for human viewing anyway (due to random phase-component)?
Deburst and merge can output complex data. What other processing is being done that may force the data to intensity - calibration, multilook, terrain correction etc?