We assume background light is additive and constant (think: dark frame)
The POLCAL task type contains over 100 images whose only difference is that they have been multiplicative scaled from one another.
Thus, if we individually normalize all of the POLCAL task type images they should be identical because the normalization will have removed the multiplicative scaling.
Any measured deviation between these normalized images is therefore caused by an additive factor AKA background light.
The new task BackgroundLightCalibration computes a separate background light measurement for each beam. This intermediate object is then used to correct solar, polcal, and observe frames in exactly the same way as we use dark frames.
We don’t correct for background light in lamp or geometric tasks. In the Lamp case it’s because a) the lamp frames are taken at a very different exposure time (which affects the magnitude of additive signals), and b) the lamp is illuminated very differently and tests show it does not exhibit the same background signal. This is OK, though because the lamp images are high-pass filtered so any background signal will be removed that way.
The geometric task is a similar story; it doesn’t need background correction because all of its data are high-pass filtered prior to analysis.
A quirk of this method is that the POLCAL data are used for background light calibration. This means that intensity-only OPs can’t be corrected. In this case we just pass through. Thus the data will be polluted with any background signal. This is a known deficiency and something and PI’s will need to know about.
This PR adds the residual background light calibration/correction to ViSP. Here’s a Confluence page going into more detail of the algorithm, but it can (hopefully) be summarized thusly:
We assume background light is additive and constant (think: dark frame)
The POLCAL task type contains over 100 images whose only difference is that they have been multiplicative scaled from one another.
Thus, if we individually normalize all of the POLCAL task type images they should be identical because the normalization will have removed the multiplicative scaling.
Any measured deviation between these normalized images is therefore caused by an additive factor AKA background light.
The new task
BackgroundLightCalibrationcomputes a separate background light measurement for each beam. This intermediate object is then used to correct solar, polcal, and observe frames in exactly the same way as we use dark frames.We don’t correct for background light in lamp or geometric tasks. In the Lamp case it’s because a) the lamp frames are taken at a very different exposure time (which affects the magnitude of additive signals), and b) the lamp is illuminated very differently and tests show it does not exhibit the same background signal. This is OK, though because the lamp images are high-pass filtered so any background signal will be removed that way.
The geometric task is a similar story; it doesn’t need background correction because all of its data are high-pass filtered prior to analysis.
A quirk of this method is that the POLCAL data are used for background light calibration. This means that intensity-only OPs can’t be corrected. In this case we just pass through. Thus the data will be polluted with any background signal. This is a known deficiency and something and PI’s will need to know about.
Â