BPP is a pretty complex but well written script.
Some people pretends that you should never use it because you lose control (but then what do they not do all their preprocessing with PixelMath?), other use it blindly with various rule of thumbs without checking the log or understanding the operations.
IMHO, like many thinks in PI, it is applicable in some cases and not in others. I see two main use cases or type of users:
- Experienced users usingit it to make fast calibration, as it handles the operations by filter automatically (I suppose that they have a library of master and undestand what happens).
- Relatively novice or casual users that want to get something reasonable out quickly.
The experienced users may need more functionality. The more casual users would prefer simplicity but, above all, robustness. The BPP may not support all manual capabilities, but the user should be informed clearly (as much as possible) when there is a problem. BPP does already a good job with its warnings, but more could be done. Currently it is difficult to recommend it to beginners because if it is not adapted to their needs they may get bad results and not be aware of the problem.
Functionality:- PEDESTAL: I am not sure that PEDESTAL is that useful if the darks are calibrated only during integration of the lights (I would need to check the source).
- Flat darks: BPP already handles dark flats if they are present. The problem is that it does not handle correctly the more common case of not having dark flat. It also do useless bias calibration of darks when we do not use optimise, but this has no functional impact.
- Master library: Maybe a simpler way to handle a 'master library', providing the root directory of such a library and letting BPP use the best dark/bias without having to load them explicitly would be useful and not too difficult (most of the logic is already in the script).
Robustness:It is not always obvious that something did not work as expected. It would be nice if BPP could summarize the processing done and the warning at the end (in the log or popup), as examining a very long log is not very practical. The kind of unusual values or results that could be checked are:
- Very small/large factors if optimization is used.
- Unusually high rejection values when building a master dark/bias
- Negative values when calibrating or integrating (light darker than the dark....)
- Abnormal median of the bias,dark,flat (for example if the flat is less than 0.1 - it is probably the wrong file).
- Abnormal relationship between those medians (bias higher than dark)
- Files ignored (for example when a file add a zero size)
- Others ?
Some of these capabilities may need the collaboration of the ImageCalibration and ImageIntegration tools (that are open source as well).
So you know what to do when it is raining
-- bitli
