In order to get the highest SNR from the data I tried different image integration parameters and compared the results. (My general set-up: non cooled Canon 600D, light frames ISO 800, 360 s at 14 °C; 40 dark frames ISO 800, 360 s at 16 °C; MasterBias from 40 bias frames ISO 800, 1/4000 s frames at 20 °C; DSLR_RAW: Pure Raw, Raw CFA; no flat frames). Best results were obtained with the following calibration adjustment:
1) Integration of dark frames, NO subtraction of MasterBias (a subtraction of MasterBias prior to the calibration of light frames resulted in severe clipping of data),
2) Calibration with MasterDark (Calibrate: checked, Optimize: checked) and MasterBias (Calibrate: not checked).
Well, every sensor contains sensor pixels that cannot provide useful information, resulting in "hot pixels". My question concerns such defective pixels and the calculation of the scaling factor k0:
Are the data from defective pixels used in the calibration process when calculating k0?
If this is the case, it ought to be changed, because the defective pixels deviate the most from linearity. My proposal:
The normal workflow assigns for cosmetic correction as the next step after calibration. I propose to evaluate a defect list (or defect map) already at the stage of calibration, that is: set the difference (MasterDark - MasterBias), that is computed during calibration, to ZERO for defective pixels. In this way it would be avoided that the calculation of the scaling factor k0 is falsified by data from defective (non-linear) sensor pixels.
Bernd
