Rev198: Saturn wide-angle mosaic Options

[wildespace]

The mosaic is the Image of the Day at NASA: http://www.nasa.gov/mission_pages/cassini/...rn20131017.html

You're a celebrity in space circles now

I like doing RGB stacks from Cassini, but I don't know how to calibrate colours (rather than adjust them arbitrarily until they resemble "true colour"), so my Titan appears white with vivid blue haze, and my Saturn is bluish-yellowish-grey. But Saturn's atmospheric features are more prominent.

[remcook]

Today in the paper. Not bad at all Gordan!
(might need to mirror the image )

MOD NOTE: Edited to remove political comment.

Attached thumbnail(s)

 

[dilo]

Congratulations, Gordan... breathtaking image!

[Ian R]

Ian, I simply assumed the original data used 12bit encoding mode used so the JPEG DNs are basically linear (a reasonable, but not 100% accurate assumption when uncalibrated images are concerned) as it looked that way to me and the color profile I use in Photoshop (and the one I use for calibrated data that is linearized by default) automatically adjusts that to the sRGB 2.2 display gamma.

Gordan,

Many thanks for the explanation: the first thing I indeed noticed was the difference in gamma between your composite and my attempt at the clear-filtered mosaic. With regards to the JPEG DNs, how would you tackle a raw you suspect of being an LUT-converted 8-bit image? Trial and error? Or is there a way to approximately reverse the conversion?

Thanks again.

[ugordan]

I haven't found a reliable way to handle the LUT-encoded stuff which makes up for the majority of Cassini imagery. It's mostly trial and error. I found it's more difficult to do mosaics as well. Theoretically, the LUT uses a square root encoding slope so if you wanted to "linearize" it, you might want to apply a 0.5 gamma (working in 16 bit depth at all times, of course), but I found that doesn't work out nice.

The good thing about the LUT encoding is you don't really have to linearize it, it's pretty close to an inverse of the default sRGB gamma so simply loading the raws into RGB channels in any vanilla photo editor and you're pretty much good to go (assuming the histogram stretch didn't massively clip dark and bright levels).

[stephenv2]

I have not done this yet but a curves preset could be build from a LUT if someone is up to the task: Photoshop curves to LUT It would be a reverse of this process.

[ugordan]

It would be a simple enough task if it wasn't for the raw histogram stretch automatically applied on the ground. Once that's done, you have no idea what original level was mapped to Jpeg DN 255, likewise for DN 0. This is what, I think, makes matching brightnesses between separate frames and mosaic footprints in LUT mode difficult. Too many degrees of freedom involved. You have the one-size-fits-all Photoshop levels curve at your end, but the actual flight LUT curve was stretched an unknown amount by the ground histogram stretch. Convolve those two and you still end up with nonlinearities.

For 12 bit stuff, especially if it shows some black space so the stretch leaves it alone, you basically have just one thing to play with - brightness.

[stephenv2]

I was under the impression that the amount of stretch was in extra lbl calibration data - or am I confusing this with SDO? We've been working with so much data lately. Kevin McAbee, the other key volunteer on the film, has been working on Cassini to automate processing of the raw files. I can check with him too.

[ugordan]

There is no camera metadata available for the Cassini raw images that I'm aware of. For PDS archived stuff this is not an issue, of course.

[stephenv2]

I'm pretty sure you are right, I don't recall seeing it either, I think that is only SDO (which has it's own sets of problems and is actually much harder to deal with). I poked around in some of my data stores and don't see anything except from PDS.

[Bjorn Jonsson]

Wow... this is awesome, especially when keeping in mind that it's from JPG data and not PDS files.

[pat]

And the mosaic using the science quality images

http://photojournal.jpl.nasa.gov/catalog/PIA17474

[djellison]

Honestly - I prefer ugordan's

[tedstryk]

Yeah, the official one is, err, really yellow.

[algorithm]

Agreed, not as asthetically pleasing as Ugordans. The photojournal site linked to a couple of posts above does however say that the image 'is as human eyes would see it'. This sounds to my untrained ears like a fairly bold statement.