Reprocessing Galileo Io Images Options

[4th rock from th...]

I'll have to play around with Saturn and Jupiter images a bit, but the results I got so far seem even duller than my usual processing. I'm not entirely sure this is the right way to get accurate colors, but the above results at least show the intrinsic color the filters would look to human eyes.

From my small experience, the only really major improvement is with images generated from filters that are different from "generic" RGB but still covering much of the visible spectrum. Voyager OGB or OGV, Galileo RGV or IRGV render nicely and don't show saturated areas or predominant casts.
As for material that's originally very close to sRGB I agree there isn't much change.

[elakdawalla]

Well, I've messed about with some Galileo images and I've come to the conclusion that attempting to get the color "right" in some technical way, like transforming the coordinates in CIE space, is just silly if I'm not working with calibrated data to begin with. And it looks like the only way to calibrate the Galileo images is to use ISIS, which I don't have access to. So I think I'll probably go back to my old method, which is just to combine filters willy-nilly, and make no claims as to the "true"ness of the color, and send people to Bjorn's website if they want a more rigorous idea of the true colors of Io.

For no reason at all, here's a global view from I24, with the green and violet mixed into the blue channel as Bjorn suggests.


--Emily

[ugordan]

As for material that's originally very close to sRGB I agree there isn't much change.

Actually, I've found that even Cassini's RGB filter combo gives weird results, even though it covers the entire visible spectrum (and only that). The fact the GRN filter is actually yellow makes Photoshop unable to represent a very big color space. Here's the color spectrum from the color picker that the RED/GRN/BL1 filters apparently can represent:

As you can see, only 3 distinct colors can be represented, I suspect this is a limitation of insufficient GRN/RED filter color separation.

You don't even want to know what IR1/GRN/UV3 gives... (green Hyperion, anyone? Iapetus, curiously, appears practically the same as in ordinary RGB)

[tedstryk]

Actually, I've found that even Cassini's RGB filter combo gives weird results, even though it covers the entire visible spectrum (and only that). The fact the GRN filter is actually yellow makes Photoshop unable to represent a very big color space. Here's the color spectrum from the color picker that the RED/GRN/BL1 filters apparently can represent:

As you can see, only 3 distinct colors can be represented, I suspect this is a limitation of insufficient GRN/RED filter color separation.

You don't even want to know what IR1/GRN/UV3 gives... (green Hyperion, anyone? Iapetus, curiously, appears practically the same as in ordinary RGB)

Only on very rare occasions do I not simply colorshift. Frankly, with Io's strange coloration, I wouldn't be too trusting of anything made from a bunch of narrowband filters.

[ugordan]

Aren't Galileo's IR,G,V filters broadband?

[tedstryk]

Aren't Galileo's IR,G,V filters broadband?

I don't think so, but I could be wrong (I am at work right now but I can look it up later).

Edit: You are right, Red, Green, and Violet are broadband. I never knew that!

[elakdawalla]

Here's the full description of the Galileo SSI filters (it took me a while to track this down, so I thought it was worth quoting):

The camera was operated in eight filtered band passes from 350-1100 nm. Because of the anti-reflective coatings used on the optics of the SSI, the short wavelength sensitivity was more limited than the camera on Voyager, thus there was no analog for the ultraviolet passband on the SSI. Otherwise, the spectral coverage was superior, both in total bandwidth and resolution, to the Voyager narrow-angle camera. The eight-position filter wheel, also inherited from Voyager, consisted of three broad-band filters: violet (404 nm), green (559 nm), and red (671 nm). These broad-band filters allowed for the reconstruction of visible color photographs and were compatible with the Voyager passbands. Four of the filters were chosen to optimize performance of the SSI in the near-infrared: two for methane absorption bands (727 nm and 889 nm), one for continuum measurements (756 nm), and one to proved spectral overlap with the near-infrared mapping spectrometer (986 nm). The final filter was a clear filter (611 nm) with a very broad (440 nm) passband.

--Emily

[volcanopele]

I will post some of the more interesting ones here with the goal of rebuilding at least some of my Io website to host them in the long term.

That new site is now up and is located at http://pirlwww.lpl.arizona.edu/~perry/io_images/

[Stu]

Some beautiful images on there, thanks... especially impressed by the images showing Io as a crescent, never seen those before...

Got me thinking about how cool it would be if a modern probe took new images of Mars as a crescent, looking back at it after a fly-by, maybe en-route to a comet or something. Then they could be used to inspire a whole new generation of space enthusiasts and...

Oh, wait...

[tedstryk]

Some beautiful images on there, thanks... especially impressed by the images showing Io as a crescent, never seen those before...

Got me thinking about how cool it would be if a modern probe took new images of Mars as a crescent, looking back at it after a fly-by, maybe en-route to a comet or something. Then they could be used to inspire a whole new generation of space enthusiasts and...

Oh, wait...

Well, maybe DAWN will do it

[edstrick]

Viking Orbiter 2 got an entire Mars rotation sequence of moderately fat ?115 deg phase? images during it's approach to mars orbit insertion. The pics have never been reprocessed into a movie.. it would take some work as they are at maybe 15 or so degrees planetary rotation. One pic, press released and widely reproduced, had a structured cloud (not volcanic) plume extending off one of the 3 tharsis volcanoes.

[Bjorn Jonsson]

This is fascinating stuff.

Regarding Io, as discussed earlier in the thread, creating synthetic B from G and V results in more accurate color than using V. The main reason is that V images are far more contrasty than the B images. To determine how to do this I used Voyager B images as a guide (this is described on my webpage here). I'm convinced the resulting color is more accurate than what you get using RGV. However, there are some caveats:

* I haven't done a careful comparison of the Voyager and Galileo filters (this needs to be done for G and V).
* I haven't checked the 'real' color of these filters, for example whether the G filter is actually closer to yellow as seems to be the case with Cassini's G filter (see ugordan's messages above).
* I wasn't using calibrated Galileo images but simply 'tweaked' the brightness of the raw PDS images using publicly released color images as a guide. This isn't as bad as it sounds because calibrating Galileo images doesn't result in big changes, unlike for example Cassini (many of the Cassini images need to be scaled nonlinearly from 8 to 12 bits and dust rings and other blemishes are more obvious in the Cassini images). Now my software can calibrate Galileo images (it couldn't when I originally did this) so the accuarcy can be improved slightly.

I haven't tried sRGB yet but doubt it would be effective when you're using IR7560 instead of R. The problem is that the reddish areas brighten significantly with increasing wavelength. This can't be seen in G images but is obvious in R images. This is also the reason the Voyagers didn't 'see' this, the O images are more similar to G than R images in this respect (in other words, it's impossible to use O and G to create realistic, synthetic R). This also means that you cannot create true color images by using IR7560 instead of R. Synthetic R is needed. Using a linear combination of IR7560 and G yielded bad results in the examples I tried. In particular I wanted a 'true color' version of the well known C21 global mosaic. However, only IR7560 is available. So using G2 images (where both R and IR7560 are available) as a guide I came up with the following nonlinear formula:

R=(IR + abs(d)^s * sign(d) * 255 x w) * m

where

w=0.582
s=1.462
m=1.092279
IR=The IR7560 intensity (0-255)
d=(G-IR)/255

(255 because the intensity range is 0-255).

What's remarkable is that this resulted in an almost perfect R image. However, the values of the coefficients probably need to be modified for use with images of the Jupiter-facing hemisphere (the values here apply to the anti-Jupiter hemisphere).

A comparison using the C21 mosaic:



The version at right is more realistic in that the the red areas appear more subdued. It can be improved further using synthetic blue and by adjusting the color balance a bit.

Cool. I just played with this and before your histogram fudging, the image does look quite a bit more like Bjorn's versions [of Io with synthetic B - BJ].

I haven't tried this (sRGB colorspace processing) yet (obviously I need to try it ASAP) so it would be interesting to see this image.

[snip]
For now, I'll just leave you with a Voyager OGB image of the Great Red Spot processed with colorspace transformation (Ok, this is going very very off topic but at least is Jupiter related!). [attachment=9918:attachment]

This looks great, the color is probably more realistic than in most of the Voyager color images I have seen of Jupiter. It would be interesting to see a global Voyager 1 or 2 image of Jupiter processed this way.

[JRehling]

This is fascinating stuff.

Regarding Io, as discussed earlier in the thread, creating synthetic B from G and V results in more accurate color than using V. The main reason is that V images are far more contrasty than the B images.

Just a wacky idea here...

Presumably, the quirks of Io and narrow filters is that the spectrum of some of the terrain types is somewhat psychopathic. I wonder, then, if it would be superior to take a MIN(x, y) of G and V rather than a weighted mean of them. Or a weighted mean of the MIN and the G and the V. In other words, kill the blips that come from a sharp reflection feature. Or, try MAX if NIMS showed those terrains to have sharp absorption features -- which seems more likely.

[Bjorn Jonsson]

I managed to dig up some information on Galileo's SSI. This was originally at http://flora.tuc.noao.edu/gllssi/index.html but apparently that site is no longer available - I found these documents on my computer:

 gll_ssi_info.zip ( 155.02K ) Number of downloads: 565


This ZIP file contains several files and one directory.

I made this spectral response graph from one of the HTML files:



Not unexpectedly, a lot of color information between 400 and 500 nm is 'missing'. There is another 'gap' near 600 nm.

[elakdawalla]

I haven't tried this (sRGB colorspace processing) yet (obviously I need to try it ASAP) so it would be interesting to see this image.

Hi Bjorn, it's quite possible I was just fooling myself; in any case I got so lost trying to figure out how to save the image with the adjusted color profile that I didn't keep it. I'm encouraged by the fact that the calibration issues aren't that big a deal with Galileo images, so perhaps I'll try it again.

--Emily