Celestia Forums

Hi,

I spent quite a bit of time today working on my "ultimate"
8k Iapetus texture, based on the recently published
Cyclops cylindrical map [Dec 2005] and further reprojected
data. Note that it also implements the "Saturn shine" hires
imaging ...

Here are the main advantages over my previous
Iapetus texture:

1) Very high geometrical accuracy, since cylindrical map
projection done by "insiders" (Cyclops team)

2) Now 8k base texture (instead of 2k before)

3) Properly normalized gray shading gives more natural
computer-mapped coloration from

http://photojournal.jpl.nasa.gov/catalog/PIA06167

4) Put lots of work into very smooth color transitions
between various texture patches of different resolution and
origin...

Here is the 8k texture for download and testing:

++++++++++++++++++++++
http://www.celestiaproject.net/~t00fri/images/iapetus.jpg
++++++++++++++++++++++

A downscaled 1k image looks like this:



Here are some Celestia images as an appetizer...



The last view is from "Saturn shine" !

Enjoy,
Fridger

Beauty. I am always impressed by the quality of surface textures we have from places so far away.

Fridger, perhaps this stripped zone in your third pict a bit smoother?

ElChristou wrote:Fridger, perhaps this stripped zone in your third pict a bit smoother?

What is "this stripped zone" in my third picture?

Bye Fridger

t00fri wrote:What is "this stripped zone" in my third picture?

Here:

ElChristou wrote:

t00fri wrote:What is "this stripped zone" in my third picture?

Here:

That part is about as smooth as it can be without sacrificing surface
information near the southern end of the Saturn shine patch. Did you
look at the 8k texture? These structures near the southern transition
are somewhat affected by cyclindrical projection artefacts, yet carry
significant information. I could only make the transition still smoother
if I partially discard information contained in that zone in the layer
merging.

This I /never/ do as a matter of principle. We want to keep a
maximum of information even at the expense of slightly non-smooth
transitions. The original scientific images don't even do ANY smoothing
for presicely the same reason (yet put to an extreme).

But I found a few other areas that need some little touch-up. I'll do
that tonight before committing the texture to CVS. I will commit a 1k
(default size) and a 4k version. 8k is too much for the distribution,
unfortunately.

Bye Fridger

t00fri wrote:...These structures near the southern transition
are somewhat affected by cyclindrical projection artefacts...

Yep, this is what exactly I wanted to avoid, but...

t00fri wrote:...yet carry significant information...

Right, let's keep this zone as it is.

Hi Fridger,
I decided to give your Iapetus texture a go, but I appear to be getting some strange color artifacts when I get closer to the surface (blue and red ghosting). Is this normal, if not, do you know the reasons for this?
The following 3 pics show what Iapetus looks like for me from various distances...

from further out:

and further again:


Thanks
Phil

Chuft-Captain wrote:Hi Fridger,
I decided to give your Iapetus texture a go, but I appear to be getting some strange color artifacts when I get closer to the surface (blue and red ghosting). Is this normal, if not, do you know the reasons for this?
The following 3 pics show what Iapetus looks like for me from various distances...
...
Thanks
Phil

Phil,

yes for everything there is a price in life .

As to my coloration by computer using natural color template images,
the pro is that one can map /very precisely/ the template
colors onto another grayscale image. The con is that the
process is not possible at a full true-color level and naturally involves
some sort of "dithering". That's precisely what you discovered above.
Since the coloration can only distinguish among 256 shades of gray,
some color mixing tricks have to be used ("dithering") to generate
more color shades in terms of 8bit color. The color mixing impression
can only work if your eyes are unable to resolve individual pixels!

+++++++++++++
HOWEVER, you are anyhow NOT supposed to look at a 4k texture
from such a close distance . It just does not make sense.
+++++++++++++

Bye Fridger

t00fri wrote:As to my coloration by computer using natural color template images,
the pro is that one can map /very precisely/ the template
colors onto another grayscale image. The con is that the
process is not possible at a full true-color level and naturally involves
some sort of "dithering". That's precisely what you discovered above.
Since the coloration can only distinguish among 256 shades of gray,
some color mixing tricks have to be used ("dithering") to generate
more color shades in terms of 8bit color. The color mixing impression
can only work if your eyes are unable to resolve individual pixels!

That's OK then. I thought there was something wrong with my graphics, and I was worried I might not be getting the best in terms of resolution etc out of your texture.

From your comments above, I gather that the data returned by the spacecraft presented only an 8bit color depth expressed as 256 shades of grey. You have used dithering in order to 'enhance' the image by matching colors to the 256 grayscales, and then interpolating between the 256 levels to give 32bit color depth.
So you've tried to make it prettier to look at in Celestia, if a little less realistic.
I say less realistic, because:
1. with the amount of solar illumination available at Iapetus I am assuming it would be quite difficult for the human eye (if one could actually be there) to percieve much more than an 8bit color depth.
2. I guess you've effectively added information in terms of extra colors that wasn't represented in the spacecraft data

So your trade-off was between depicting it realistically in 256 colors, or making it a bit flashier to look at in Celestia at the expense of realism.

Thanks for the explanation.
Am I understanding you correctly?

Cassini's ISS cameras use 12bit ADCs, although many of its images do get reduced to 8bits before transmission to Earth.

See http://pds-rings.seti.org/cassini/iss/ISSNA_INST.txt

The 8bit limitation is in your computer's graphics card, which has an 8bit DAC on each of the three color channels.

Dithering is used to try to pack the 12bits of shading into the 8bits that are available.

Chuft-Captain wrote:
...
So your trade-off was between depicting it realistically in
256 colors, or making it a bit flashier to look at in Celestia
at the expense of realism.

Thanks for the explanation.
Am I understanding you correctly?

Not really, I suppose.

I use a GIMP module for doing the coloration that offers the
option "use subcolors". The latter essentially meaning
"dithering". Here are some further respective explanations:

Suppose all we got is the possibility of depositing black
dots on a white background. A laser printer for example.
Still we know that with a clever driver we may generate
pretty neat gray shades, /AS LONG AS/ we don't look at
that image from too close a distance. What the dithering
driver does is to simulate the shades of gray by distributing
the black dots at different /densities/ on the white paper!

Next, with 8bit gray scale imaging, say, a clever dither
algorithm may /emulate true 16bit gray shading/ (again
not looking too closely). The point is that neighboring 8bit
gray shades are "rough", i.e. they jump from pixel to pixel
since we only got 256 values in total to chose from. The
dithering effects some smooth interpolation of these values
by exploiting the point density degree of freedom in
addition...

The GIMP coloration routine with the option "use subcolors"
just employs a similar technique to arrive at a MUCH
improved color mapping of the grayscale original.

The price to pay are those strangely colored pixels that you
discovered at a NONADMISSIBLY high zoom level...

Bye Fridger

Chuft-Captain,

I updated the 4k iapetus in CVS with a slight tune-up that
eliminates much of this colored dithering "noise" with a
careful "selective Gaussian blur". Have a look whether this
looks better from your point of view at large zoom...

Bye Fridger

T00fri,
Have you tried using a noise filter? My favourite is noiseware from imagenomic. Its very powerful and almost seems to work like magic on certain types of noise, lifting it off while leaving the underlying image completely intact! I use it alot for digital photos. There is a free community edition on the download page below...

http://www.imagenomic.com/download.asp

I cut out a section from your 8k texture linked above and filtered it with default settings. The results are impressive...

Original...

Filtered...

t00fri wrote:Suppose all we got is the possibility of depositing black
dots on a white background. A laser printer for example.
Still we know that with a clever driver we may generate
pretty neat gray shades, /AS LONG AS/ we don't look at
that image from too close a distance. What the dithering
driver does is to simulate the shades of gray by distributing
the black dots at different /densities/ on the white paper!

Fridger,
Nice analogy, just using pixels instead of pico-litre sized droplets of ink.

Your reward for such a good explanation is these stunning photos of Mars:
http://www.geocities.com/traitorsclaw/f ... emarte.pps

DBrady wrote:T00fri,
Have you tried using a noise filter? My favourite is noiseware from imagenomic. Its very powerful and almost seems to work like magic on certain types of noise, lifting it off while leaving the underlying image completely intact! I use it alot for digital photos. There is a free community edition on the download page below...

http://www.imagenomic.com/download.asp

I cut out a section from your 8k texture linked above and filtered it with default settings. The results are impressive...

Original...

Filtered...

I dont find your two examples impressively different, I am afraid.
Of course I use all sorts of noise filters in GIMP or Photoshop if I have to. There is surely no need for "filter magic" from some other firm. But thanks anyway for the link.

As I wrote above your post, I have now gotten rid of most of the random colored pixels by using a "selective Gaussian blur" filter (together with some channel masks etc.) which despite its name, usually effects a sharpening rather than blurring if the parameters are set correctly.

Bye Fridger

Chuft-Captain wrote:

t00fri wrote:Suppose all we got is the possibility of depositing black
dots on a white background. A laser printer for example.
Still we know that with a clever driver we may generate
pretty neat gray shades, /AS LONG AS/ we don't look at
that image from too close a distance. What the dithering
driver does is to simulate the shades of gray by distributing
the black dots at different /densities/ on the white paper!

Fridger,
Nice analogy, just using pixels instead of pico-litre sized droplets of ink.

Your reward for such a good explanation is these stunning photos of Mars:
http://www.geocities.com/traitorsclaw/f ... emarte.pps

Thanks for the images, but why did you send them in such a complicated manner packed into Powerpoint?? Where are they from? If these are Celestia images, I can recommmend looking up some of mine

Bye Fridger

http://www.celestiaproject.net/~t00fri/images/mario32k.jpg
http://www.celestiaproject.net/forum/viewtopic ... light=mars
http://www.celestiaproject.net/~t00fri/images/MarsM31.jpg
http://www.celestiaproject.net/~t00fri/images/mm0.jpg
http://www.celestiaproject.net/~t00fri/images/mm1.jpg
http://www.celestiaproject.net/~t00fri/images/mm2.jpg

t00fri wrote:...I dont find your two examples impressively different, I am afraid...

Fridger, just for info, perhaps your king size screen has pixel too small to appreciate the difference, but yes the colored pixels have been eliminated in the second shoot. the difference is quite visible on a 1024...

That's possibly a very good point El Cristou,

I think the effect / impact of any sort of artifact in jpegs, may depend on the resolution at which it is viewed, as the mapping of image pixels to physical pixels will be different if the resolution of the screen it was created on, and the one on which it is viewed on is different. It would be interesting to know what Fridger's screen resolution is.
I generally view on a 1024x768 laptop or 1280x1024 flat panel monitor.

I wonder ....

ElChristou wrote:

t00fri wrote:...I dont find your two examples impressively different, I am afraid...

Fridger, just for info, perhaps your king size screen has pixel too small to appreciate the difference, but yes the colored pixels have been eliminated in the second shoot. the difference is quite visible on a 1024...

Right. That just illustrates the idea of dithering . It all works fine as long as single pixels cannot be resolved, which I indeed cannot on my screen...

Anyhow I would appreciate if someone could inspect my new 4k iapetus texture that I re-committed to CVs yesterday. The colored pixels are also /much/ reduced.

Bye Fridger