Celestia Forums

Hi all,

My new Saturn rings are now ready. I've used 6 raw images from Cassini to build a true colore picture and extract a 2k ring texture. A little problem was the alpha channel. Grant's alpha texture can't be used. After some calculations and tests it was generated by splitting the base texture into ring color and ring transparency.

Now some screen shots:


A compare with a true color Cassini picture.


Parts of the B-ring, Cassini divission, the A-ring with Encke and Keeler gap, the faint F-ring and nearby Atlas and Pandora.


The dark C-ring.

For all shots was futher used my Saturn map.
Both the new rings and Saturn map can be found on my Site.

Have fun!
Jens

jim wrote:Grant's alpha texture can't be used.

Hi Jens:
Nice rings!
For interest's sake, what was the problem with combining my alpha channel with your ring colour channels?

Grant

Jens,

Just wanted to Thank You for providing the new rings textures.
Absolutely beautiful. Thanks also for providing the 1k version for us low-tech
types. Now, I'm going to have to revise my Saturn Tour. Again, beautiful
work. Thank you sir.

Take care, Bob

granthutchison wrote:For interest's sake, what was the problem with combining my alpha channel with your ring colour channels?

Hi Grant,

There is a little but significant missalignment between your alpha channel and the new texture. Here you can see what I mean:

http://www.celestiaproject.net/~jim/images/comp_rings.png

Jens

Hi Jens,

really, your new rings are very nice and I use them already since yesterday! What I noticed is, however, that the generic transparency of your rings is less than the one Grant has (carefully) determined from the earthbound transit data. I have a fairly good feel for them, since I went "closely along with Grant's analysis"...How much would you want to insist on your (average) transparency level?

Bye Fridger

Wouldn't a quick test of the transparency channel be to compare Celestia's ring shadows on Saturn with the Cassini pictures?

I haven't done this yet, but I did notice that the shadows cast by Jens' rings are noticably different from the shadows cast by Grant's.

granthutchison wrote:To improve my own ring texture, I've combined raw images 349-351 into a colour composite, toned down the contrast and saturation, and then sampled across the B ring to extract some albedo and colour detail. I've then added that to the RGB channel for the B ring while leaving transparency untouched - it certainly gives a more realistic texture to the otherwise rather bland areas produced by low transparency.

OK, Selden has now made my revised 4k rings available if anyone wants to take a look: http://www.lepp.cornell.edu/~seb/celestia/hutchison/saturn-rings.html. The only difference is in the B-ring, but it's a noticeable improvement.

Grant

PS: Oops. Just edited my previous post instead of quoting it, so it has now vanished from the thread. .
Apart from the bit I quote above, the only significant loss from my prior ramblings is my reply to Selden's comment - since Chris fine-tuned the shadow density using my rings, it's probably a bit of an unfair test for Jens' texture.

Fridger wrote:What I noticed is, however, that the generic transparency of your rings is less than the one Grant has (carefully) determined from the earthbound transit data. I have a fairly good feel for them, since I went "closely along with Grant's analysis"...How much would you want to insist on your (average) transparency level?

Fridger, I use a very simple methode to generate my ring transparency. Therefore I would absolute not say that they are accurate. I think that parts of the B-ring are to transparent.

Fridger, I know that my alpha channels is not accurat but I know further that the result (the combination of alpha and rgb channel) "appears" very accurate in Celestia. The question is what's to prefer an accurate alpha channel (Grant's one) or an accurate appearance (the product of alpha and opaque channel)?

I think I should explain how the alpha and opaque channel have been split from the (raw) rings. First I changed in PSD to Lab color mode to separete albedo and color information. Then the albedo channel was used to divide into alpha and opaque channel by this simple equation (in PSD with gradation curves):

rings = opaque channel * alpha channel

I'd tested several solutions:

1. opaque channel = square root (raw rings)

2. opaque channel = 1 - 0,25 * raw rings

3. opaque channel = 1 - 0,5 * raw rings


I'd compared these with Cassini pictures and came to the result that last methode gets the best results. But as you can see there is a lot room to improve.

I will make some experiments at the weekend with Grant's alpha channel maybe I find a way to use it. But I have still some doubts about the accuracy of Grant's alpha especially the faint parts of the C-ring.

Jens

Grrrr !!! Foreget to login and can't edit.

Jens

jim wrote:Grrrr !!! Foreget to login and can't edit.

Jens

Jens,

I really feel that the overall appearance of your rings matches the photographic images amazingly well.

Admittedly better than the ones by Grant...

However, the photos only display particular views and thus, from such a comparison, we can get at best partial information about the ring transparency. Personally I consider the transparency issue very important, since indirectly it also influences the conspicuous ring shadow patterns considerably etc.

Unfortunately I am presently very busy. But I hope I'll find some more time over the weekend. Then, I would be very keen to go on discussing this important issue with you and Grant!

Bye Fridger

jim wrote:The question is what's to prefer an accurate alpha channel (Grant's one) or an accurate appearance (the product of alpha and opaque channel)?

If the alpha channel isn't accurate, you'll never convincingly model ring shadows (Chris can fiddle the shadow density, but we know from early experience that a transparent ring casting a dark shadow looks wrong), and you'll never model the correct appearance of Saturn's disc seen through the rings (which we can check against photographs). So "accurate appearance" does depend very strongly on an accurate alpha channel in many instances.
You also have to beware the stretched contrast routinely used in ring photographs, which creates the false appearance of bright lanes and black gaps in opaque rings, and pulls out detail of very faint structures, making them appear brighter and denser than they actually are. Remember our debate over the appearance of the F-ring?

Grant

Hi all,

Fridger wrote:Personally I consider the transparency issue very important, since indirectly it also influences the conspicuous ring shadow patterns considerably etc.

I don't say that the transparency issue is unimportant. I will furhter try to improve my rings in this case. At the moment I don't pay to much attention to the shadows because they are also affected by Celestia.

granthutchison wrote:If the alpha channel isn't accurate, you'll never convincingly model ring shadows

Grant, I never tried to adjust my rings for correct shadows because this should be the last step and can't done by me.

granthutchison wrote:(Chris can fiddle the shadow density,...)

These parameter in 'celestia.cfg' would be very nice.

granthutchison wrote:and you'll never model the correct appearance of Saturn's disc seen through the rings (which we can check against photographs).

I agry and as you metion "check against photographs" that's exact what I do - very, very carfully.

granthutchison wrote:So "accurate appearance" does depend very strongly on an accurate alpha channel in many instances.

Not in front of a black backround! For this case I can say my rings appears nearly exact.

granthutchison wrote:You also have to beware the stretched contrast routinely used in ring photographs, which creates the false appearance of bright lanes and black gaps in opaque rings, and pulls out detail of very faint structures, making them appear brighter and denser than they actually are.

Grant, I see you think similar to me "before" I build this rings and looked at all those Cassini picturures. I thought bright lanes and black gaps in opaque rings aren't possible. These structures can be only caused be transparency of the rings. But look at this picture:



As you can see the dark "gaps" of the B-ring don't change there apeariance in front of Saturn or the black space. Further we see that the inner part of the B-ring lock very similar in color and transparency to the A-ring. The C-ring and especialy it's shadow dosen't look so nice because the raw images are overexposed. Nevertheless we can see that the matierial of this ring is dark but not too dark to appear still bright in front of the black space.

Before I forget I'd adapted the albedo of my rings to the latest complete Saturn picture to get the correct colors and avoid unnatural contrast.

* * *

Grant, from the scientific point my work may not accurate but you should know I try always my best.
I'm always open for a discussion that help something to improve. My aim is that all should appear as real as possible in Celestia and when this can be done with a bit (unscientific) artwork - why not?

Jens

Thanks, Jens, that's a beautiful image. But I'm beginning to wonder if we aren't at cross purposes, here, since the image seems to completely support everything I said in my last posting, and confirm the correctness of my upgrade to my own rings:
1) The dark lanes are albedo features of the B ring, not transparency features.
2) The B ring is completely opaque in its central region, as it appears in my alpha channel.
3) The dark lanes are in fact relatively bright.

The only point Fridger and I are trying to make is that your data extraction method makes these dark lanes into regions of transparency, which you've just shown with your photograph isn't correct. If you check your own ring texture against the bright limb of Saturn, you'll see that the B-ring does change brightness as it crosses the limb, because it is transparent enough to let Saturn show through.

Grant

Hi Grant and Fridger,

I've built these diagrams to show the differences between our alpha channel.



I was surprised how "less" transparent my B-ring still is. If you realy think they should be more opaque I will do it. Let me know if you see other things to improve.

Jens

Hi all,

here are some more pictures to study the rings.


link to the highres version: http://www.celestiaproject.net/~jim/images/cass_rings5.jpg


The official released picture PIA06060 but a bit color adjusted by me.


Enhanced gamma to make the transparency of the B-ring visible.

Interesting: there are three signifant opaque parts.
1. the centre of the B-ring,
2. the last bright rings near to the inner more dark strip of the B-ring,
3. a dark ring at the inner end of B-ring.

Jens

Hi all,

it is done! A improved version of the "real color" Cassini-Saturn-rings is ready. I didn't know how expensive this job could be. The idea was to combine Grant's alpha with my rings. The main problem was I must notice that something is wrong with Grant's alpha. Parts of the A-ring, the transparent gaps of the B-ring and the faint parts of the C-ring are to transparent. On the the other hand the position of these tranparent features are very exact. All tries to adapt Grant's alpha was without success. Now I search for another methode to get an better transparency for my rings. The problem is that we have one equation with three variables while only one is known - the raw rings (from the Cassini pictures) as product of tranparency (alpha channel) and material albedo (opaque channel). Therefore I had to collect some additional informations in combination with some assumptions.
1. the matierial of the C-ring is relative dark,
2. large parts of the B-ring are opaque, that means the the structures of the B-ring are mostly given by albedo features,
3. the structures of the less opaque rings are mostly given by the transparency,
4. the material is mostly ice and should by general relative bright.

Now my solution:
I've decided to build the opaque channel (albedo of the material) first and then use this to calculate with the "raw" rings the alpha channel (transparency). As base I used my "old" opaque channel and modifiied this partly manually and applied partly a lowpass filter. The C-ring was generated by useing a modified version of Grant's alpha. After a lot calculation with Excel my new opaque channel was ready. Now the rest is easy only a quick calculation with Excel and the new alpha channel was created.

Finally some diagramms to prove the results:


Before I foreget a better texture for the F-ring was added. But except this the only difference to my previous version is the modified transparency. That means you must look very carfully to find the differences between my rings because the general appearance has not change. I think someone can now better unterstand what I mean with "appearance in Celestia".

The new rings can be found on my homepage.

Any comments or suggestions are wellcome!

Jens

P.S. 1. I don't know if "opaque channel" is the correct expression. Maybe somewhere can enlighten me?
2. If somewhere has an interest on my Excel table (ca. 3mb) or the alpha and rgb cannel as single files I can upload these.

jim wrote:The main problem was I must notice that something is wrong with Grant's alpha. Parts of the A-ring, the transparent gaps of the B-ring and the faint parts of the C-ring are to transparent.

Jens, that alpha channel is real, measured transparency - it's the absolutely best data we have. It doesn't match the ring photographs because the ring photographs are enhanced to bring out faint detail, so of course the faint parts of the rings look more dense than my alpha channel makes them.

But we've had this exchange three times now, and although you seemed to eventually accept the argument with reference to the F-ring, you obviously don't believe it applies elsewhere. So, since it's probably getting annoying for all concerned for us to go around this over and over again, this'll be my last post concerning Saturn's ring transparency.

Grant

granthutchison wrote:

jim wrote:The main problem was I must notice that something is wrong with Grant's alpha. Parts of the A-ring, the transparent gaps of the B-ring and the faint parts of the C-ring are to transparent.

Jens, that alpha channel is real, measured transparency - it's the absolutely best data we have.

Yes, in principle. I am sure you recall a discussion of the systematic inherent uncertainties from the original publication. I remember, I glanced through it when you did your analysis, but unfortunately I don't recall all the sources of error anymore...

My guts tell me though that there should be quite a few uncertainties that we have to fold into that discussion.
The displayed measurements and the apparent resolution look clearly impressive. But there were no error bars ...

Bye Fridger

Ah, Fridger, I'll e-mail you.
I've posted everything I think about this topic three times already - everyone (including me) deserves a rest.

Grant

hi,you post some very nice screen shot here!
I notice that your saturn ring can drop shadow on the surface of the planet.But my saturn dosen't have the shadow of the ring ,can you tell me how to get the same effect as yours.