I'm working with some New Horizons MVIC/Ralph data. I'm taking the red, blue, and NIR channel data and create a false colour image, like this: (PDS Label of Red Channel image, accessed from OPUS)
Because of the nature of the MVIC/Ralph instrument (offset long, skinny CCDs in a time-delay integration scan) the individual channels need alignment. I manually aligned the above image but I am looking to automate/script this process.
Luckily, the angular offset of each channel is known:
I tried to use this information to calculate how many pixels to move each layer for a different image (PDS Label). The image is 5024 x 7776 px and page 6 of Reuter et al. gives the single pixel FOV as 19.77 µrad. This translates to ~50582 px/rad. Each channel is separated by 0.037° (19.77 µrad x 32 px) + 0.070° $\to$ 1.854 mrad. I expected this to mean having to move each channel by ~94 px (50582 px/rad x 1.854 mrad). This is short by about 10-12 px still:
When applied to subsequent channels (in multiples) the 10-12 px offset remains constant.
What other factors could be at play here? I don't think it's the 12 dead pixels at the edge of each CCD as those are in the "other axis," but I could be wrong.
- Reuter et al. "Ralph: A Visible/Infrared Imager for the New Horizons Pluto/Kuiper Belt Mission," Arxiv.org Link