Difference in calibration between ISIS and OPUS Voyager 2 Images

[JuB104]

Hi,

I have some .imq images from the Voyager 2 fly-by of Uranus which I am processing using ISIS. I apply voy2isis and spiceinit to them before calibrating them using voycal. Now I noticed that the max I/F that results from this process is different from the one that is obtained when using already processed images from the OPUS3 website (https://opus.pds-rings.seti.org/), especially when using the geometrically calibrated 1000x1000 images.
Is it possible that ISIS' voycal is performing a different radiometric calibration than NASA's Planetary Data System?

Thanks in advance.

[dpmayerUSGS]

The derived products available from OPUS, such as geometrically corrected Voyager ISS images were processed by the Multimission Image Processing Lab at JPL, not the PDS (the derived products are also technically non-canonical). MIPL used the VICAR software to create those products using a custom pipeline, so it wouldn't be surprising if the calibrated results from MIPL don't match the results of processing the same raw product in ISIS. It's difficult to say which processing pipeline is "better."

Information about MIPL's workflow is available in the processing notes file in any of the Voyager ISS volumes in OPUS, such as https://opus.pds-rings.seti.org/holdings/volumes/VGISS_7xxx/VGISS_7205/DOCUMENT/PROCESSING.TXT

[blandoplanet]

I'm not 100% sure what the MIPL software does, but a similar correction can be performed in ISIS using the findrx application. This uses pattern matching to locate and refine the position of the reseaux. The reseaux position information is then overwritten in the attached cube label and so is then propagated into the geometric distortion correction any time the camera model is used. The change can be significant (depending on the image). To test this, take two copies of the same image, run findrx on one of them and then compare the output from campt. In my view, it's always best practice to run findrx when using Voyager data, since the distortions can be large.

[dpmayerUSGS]

I concur with @blandoplanet regarding findrx.
spiceinit should be run on the geometrically uncorrected images.

[JuB104]

Thanks @blandoplanet for the answer, I missed the part in the ISIS documentation where it was explained that findrx corrects the geometric distortion.

Also Thanks @dpmayerUSGS. So if I understand correctly SPICE does not account for the distortion of Voyager's cameras but produces data for the 'raw' images/ distorted camera model?

[dpmayerUSGS]

Ah, we may be using "geometrically uncorrected" to mean different things. findrx does not resample or reproject the pixels in the input image. The program updates the label with corrected sample/line locations of reseauxs. If you look at the label of a Voyager ISS image before running findrx, you'll see a table of sample/line locations of reseauxs. If you look at the image in qview, many of the reseauxs in the image will not appear at the sample/line coordinates reported in the label. findrx uses a pattern matching template to try and identify reseauxs in the image and updates the table in the label.

The MIPL processing seems to be different in that they actually warped the original images to force the reseauxs into their nominal positions within a 1000 x 1000 pixel image.

A typical Voyager ISS workflow in ISIS would involve running voy2isis, findrx, remrx (if you want to remove the reseauxs), spiceinit, voycal, and then cam2map if you want to transform an image from the original camera geometry into a map projected image. spiceinit will take into account the reseaux location table in the input image, so the results should be more realistic if you run findrx before spiceinit.

[JuB104]

Okay Thanks, that makes sense! I will try out the pipeline you described.