What is the mathematics of the adaptive pedestal?

Interesting question, I've forwarded it to Mabula.

It's quite puzzling. The documentation says that it only affects dark frame calibration. But when I have the option on it sets the adaptive pedestal for the master dark to zero and it only affects the light frames. This seems to be a reversal of the documentation.

Hi @ddnum & @vincent-mod,

The adaptive pedestal will only kick in / work if darks/masterdarks are available. And it will do nothing with the darks but it will affect the dark subtraction of the lights.

If you have under-exposed your light frames and a MasterDark is subtracted, many pixels will be clipped to the black point / 0 value.

Now the adaptive pedestal will correct for this is this happens. If no pixels are clipped to 0, the adaptive pedestal will do nothing on the light frame calibration. If a significant number of pixels is clipped to 0, the adaptive pedestal will kick in.

When it kicks in, the light frames will be modified by adding an artificial pedestal, so less pixels will be clipped when the MasterDark is subtracted. On seriously under exposed light frames, the adaptive pedestal will still have the effect that still no pixels will be clipped to zero 😉

I checked the explantion of the Tooltip for the adaptive pedestal and it is correct and explains why this is important for correction of severe Amplifier Glow in the dark current of sensors.

If you have different images that have different adaptive pedestals applied, it means that under-exposure of your frames has changed. The adaptive pedestal will start with adding Y ADUs where Y ADUs actually is derived from the MasterDark statistics. If Y ADUs still lead to clipping to zero for many pixels, 2xY will be added, if that still is not enough, 3xY will be added 😉

So that should help you to compensate for your moonglow statistics.

Mabula

Thanks. This makes sense. But I don't understand why it is useful for amp glow. If the darks have amp glow then the lights will have amp glow so when you subtract the dark from the light you shouldn't get negative values.

@ddnum, as explained by the tooltip, if the light frames are underexposed then subtracting a MasterDark can lead to clipping on the black point/0 value, right?

Now, with severe amp-glow and these new CMOS sensors, actually many pixels will start to clip in that amp-glow region specifically... And if you don't use an extra data pedestal/offset, the pixel integration values get skewed in that amp-glow area to the upside, because the pixel value distribution in the pixel stack is cut off on the low side... leading to visual amp-glow 😉 dus to skewed statistics.

I did an integration and it turned out to have the same pedestal as the individual calibrated frames. This is the histogram of the integration - the pedestal is 0.11

 Two questions:

1) to use this image in other software not aware of the pedestal I should subtract the pedestal from each pixel right? I get a few negative pixels that way but I guess that's OK

2) there are a few pixels with zero value. Where could they have come from? They seem like outliers. My guess is something to do with registration and extra pixels added on the edges.

Hi @ddnum,

The pedestal of the integration is the same as the pedestal of the reference ;-).

Indeed, you can subtract the reported pedestal of the integration file from all pixels in the integration. Some pixels will be clipped, yes, that is why the pedestal has kicked in 😉 If there was no clipping, no pedestal would have been added.

Of course, some zero pixels could be on the border as well due to the full composition after registration of all files.

Mabula