IPB

Welcome Guest ( Log In | Register )

 
Reply to this topicStart new topic
Color images from spacecraft images and a spectrum, Image processing experiments
Bjorn Jonsson
post Jun 2 2013, 12:31 AM
Post #1


IMG to PNG GOD
****

Group: Moderator
Posts: 2250
Joined: 19-February 04
From: Near fire and ice
Member No.: 38



In a different thread where I posted recently processed Voyager 2 images of Europa I mentioned an image processing idea: Somehow combining Europa's entire visible spectrum with color information from spacecraft images to compute a synthetic, visible spectrum for each point in an image and use this to make synthetic R/G/B images.

I experimented a bit with this and wrote a small program to do this - the result was successful:

Attached Image


This image was generated by combining Europa's visible light spectrum from 380 to 770 nm with data from orange, green, blue and violet filtered Voyager 2 images. I used calibrated images from the PDS Rings Node. What I did was to use Europa's spectrum instead of linear interpolation to create a synthetic spectrum that is then converted to sRGB. It's illustrated in this graph:

Attached Image


Everything is done with an interval of 10 nm, this is also the reason the Europan spectrum appears rather smooth.The spectrum I used is of Europa's leading hemisphere which is visible in the right half (and actually extends a bit into the left half) in the image above. The Voyager 2 data and the synthetic spectrum is in arbitrary units.

The Voyager data represent an average from a north-south intensity profile from Europa's leading hemisphere in the Voyager 2 images I used. Interestingly, the green filter value from Voyager 2 is clearly too high and I'm pretty sure this is real since compared to the spectrum, it's too high relative to OR, BL and VI almost everywhere in the source images. A possible reason is that the Voyager filters are broadband filters; even though the green filter's effective wavelength is 564 nm it is sensitive to a range of wavelengths. This can be corrected by multiplying the orange, blue and violet data with ~1.08 before creating the synthetic spectrum and I decided to do this even though I'm usually not a big fan of fudge factors. The image above is the result after this correction. If I omit this correction the resulting image looks rather green - almost certainly too green.

Looking at Europa's spectrum it is clear that this method should be more accurate than using linear interpolation between the Voyager filters. In particular, linear interpolation underestimates the amount of green since a straight line between the blue and green Voyager 2 data points lies below Europa's spectrum everywhere with the exception of the data points (the effective wavelength of the Voyager 2 filters is VI=402 nm, BL=475, GR=564 and OR=589).

The algorithm I'm using for converting the spectrum to R/G/B is discussed in this interesting thread (one of the most interesting image processing threads I've seen here at UMSF). The spectrum -> RGB conversion really is the crucial part in all this. Like Gordan I'm also using Andrew Young's color matching code. My code therefore has at least partially the same issues as Gordan's - in particular the illuminant C issue mentioned by Gordan in the first post. I'm using the correction factors near the top of this post to deal with it.

The spectrum to RGB conversion is probably the biggest possible source of error and the problem is that I'm not exactly an expert in that stuff. But *if* the spectrum to RGB conversion works correctly (and it is probably either correct or nearly so) the resulting image of Europa should be extremely close to Europa's true color - at least on a global scale.

There is some additional, interesting image processing discussion in this thread (it starts here) and also in this thread (the image processing discussion starts here).
Go to the top of the page
 
+Quote Post
ugordan
post Jun 2 2013, 04:43 PM
Post #2


Senior Member
****

Group: Members
Posts: 3648
Joined: 1-October 05
From: Croatia
Member No.: 523



I also found the green filter to be too bright and I'm not even sure if it's equally bright between the 4 different cameras. I suspect it's not an issue related to bandpass since Mimas and Enceladus appear too green as well and yet, they have fairly flat visible spectra. In fact, all the O/G/B/V channels look funky to me, when compared to Cassini (or in Jupiter case Galileo). I tried working out some correction factors from Mimas and Enceladus myself, but the results are inconsistent since the datasets are pretty poor. It appears to me it's not simply the green filter, it's the relationship between orange and green, orange and blue, etc.

Here's my quick take on that image, simple linear spectral interpolation from PDS calibrated OGB data (I noticed V doesn't contribute much when you have B so I didn't bother) and using the following correction factors: 0.96*O, 0.93*G, 1.02*B:

Attached Image


On the left is a gamma-correct representation, on the right a version similar to yours.

Personally, I wouldn't trust Voyager-derived color much. The filter ranges are limited and the calibration is at odds with later missions.
As far as color matching code is concerned, I moved on to simply integrating over the entire spectrum the XYZ components multiplied by source spectra and D65 illuminant and then just converting that to sRGB with no correction factors applied later. When I switched to this method in VIMS code, the results looked pretty much the same as the old method.


--------------------
Go to the top of the page
 
+Quote Post
ugordan
post Jun 6 2013, 01:17 AM
Post #3


Senior Member
****

Group: Members
Posts: 3648
Joined: 1-October 05
From: Croatia
Member No.: 523



I've been looking into this some more. First, I take back the comment the violet filter doesn't affect Voyager color rendition, it's not a huge effect, but is noticeable. Mostly in bringing ice color down toward more reddish than compared to simple OGB spectral extrapolation.

Second, I'm even more convinced the calibration uncertainties for even the single physical camera are noticeable. There appears to be some residual nonlinearity in DN response so an image that produced a low signal in one exposure could look noticeably differently with a longer exposure after calibrating to same physical units of reflectance. Here's an example of Io, same green filter images taken in succession, but with varying exposure (images greatly reduced in size).
Attached Image

Notice the difference in Jupitershine. I worked out a 10% brightness discrepancy right there. Given this kind of thing it's no wonder we need additional correction "fudge" factors.


--------------------
Go to the top of the page
 
+Quote Post
Bjorn Jonsson
post Jun 7 2013, 12:22 AM
Post #4


IMG to PNG GOD
****

Group: Moderator
Posts: 2250
Joined: 19-February 04
From: Near fire and ice
Member No.: 38



I'm not surprised there are brightness discrepancies, I have noticed these when working with underexposed/dark images but I haven't measured them yet. They could even be a flatfielding problem (which itself might be a consequence of residual nonlinearity). Also the image corners and a fairly large area near them are a notorious source of 'bad' data, even when using the calibrated images.

QUOTE (ugordan @ Jun 2 2013, 04:43 PM) *
Here's my quick take on that [Europa] image, simple linear spectral interpolation from PDS calibrated OGB data (I noticed V doesn't contribute much when you have B so I didn't bother) and using the following correction factors: 0.96*O, 0.93*G, 1.02*B:


Nice and the version at right is remarkably similar to mine - the color difference isn't big. Needless to say I don't know about Europa's exact color but I think I prefer the color in your version.

I have been experimenting a bit more and I have found that I also need a green (GR) correction factor when using Jupiter's global spectrum to spectrally interpolate the OGBV data. I get a great looking color composite of Jupiter from Voyager 2 data using the same correction factor (0.93*GR) as I used for Europa but if anything I need to darken the GR image even more.

As a sanity check I also decided to use Saturn's global spectrum to determine Saturn's global color since I know what to expect there. I got this:

Attached Image


The brightness is at maximum at left (brightest color channel=255) but at right it has been multiplied by 0.75. The color is similar to what I was expecting and hoping to get but as usual it looks rather saturated. Applying gamma correction results in something very similar to what I've seen through a telescope although any comparisons to that are difficult and highly subjective.

For Europa I got this by using the spectrum of the leading hemisphere:

Attached Image


In addition I got a nice blue color for Uranus and white (very slightly bluish) for Enceladus but the Enceladus spectrum I have is rather crude but I should be able to find a better one. In fact I'm still trying to track down better spectra than the ones I have for several objects.
Go to the top of the page
 
+Quote Post
scalbers
post Dec 14 2014, 03:51 PM
Post #5


Senior Member
****

Group: Members
Posts: 1621
Joined: 5-March 05
From: Boulder, CO
Member No.: 184



I've been reading and experimenting with color spaces and the like recently, so I though it would be good to include some basic articles like this:

The CIE XYZ and xyY Color Spaces

As mentioned elsewhere, the question of what to do if the sun's color is different from the white point, and whether chromatic adaption should adjust this is an interesting judgement call.


--------------------
Steve [ my home page and planetary maps page ]
Go to the top of the page
 
+Quote Post
scalbers
post Jan 1 2015, 10:39 PM
Post #6


Senior Member
****

Group: Members
Posts: 1621
Joined: 5-March 05
From: Boulder, CO
Member No.: 184



I'm getting improved results with renderings of Earth's clouds if I use a D50 white point in in the XYZ to sRGB transformation. Then I set the computer display to a brightness temperature of 5000K. Fewer issues are apparent with chromatic adaptation.


--------------------
Steve [ my home page and planetary maps page ]
Go to the top of the page
 
+Quote Post

Reply to this topicStart new topic

 



RSS Lo-Fi Version Time is now: 28th March 2024 - 07:38 PM
RULES AND GUIDELINES
Please read the Forum Rules and Guidelines before posting.

IMAGE COPYRIGHT
Images posted on UnmannedSpaceflight.com may be copyrighted. Do not reproduce without permission. Read here for further information on space images and copyright.

OPINIONS AND MODERATION
Opinions expressed on UnmannedSpaceflight.com are those of the individual posters and do not necessarily reflect the opinions of UnmannedSpaceflight.com or The Planetary Society. The all-volunteer UnmannedSpaceflight.com moderation team is wholly independent of The Planetary Society. The Planetary Society has no influence over decisions made by the UnmannedSpaceflight.com moderators.
SUPPORT THE FORUM
Unmannedspaceflight.com is funded by the Planetary Society. Please consider supporting our work and many other projects by donating to the Society or becoming a member.