Just a Test -- 16k Colored Moon

[t00fri]

Hi all,

unfortunately there is simply NO good Moon imaging, despite the Moon's closeness...

Until hires JAXA/SELENE (KAGUYA) images will be available, we will have to be satisfied with UVVIS/Celementine results. While the latter exist as 90k raw data and 64k VTs have been easily produced with my F-TexTools, color is missing at this highest resolution and there are plenty of artefacts.

After struggling with the 64k Clementine data for several days, I returned to the 5 band UVVIS/Clementine data that exist at 21834 x 10917 resolution such that a nice 16k colored texture set can be done with it.

http://ser.sese.asu.edu/MOON/CLEM_COLOR ... est.cub.gz
http://ser.sese.asu.edu/MOON/CLEM_COLOR ... ast.cub.gz

http://ser.sese.asu.edu/MOON/clem_color.html

The west and east hemispheres of the moon are about 1.2 GB each after decompression.

They are in ISIS2 cub format and the resolution is 500meter/pixel. The three r,g,b colors have to be extracted from the set of 5 color bands with an appropriate isis3 tool.

While this set is again full of artefacts, I think the color is quite impressive with lots of subtle shades.

To explore the "potential" of these colored data set to the end, I used much of the ISIS3 tools arsenal for thorough cleaning of the raw data. An important positive fact is that the black NULL pixel areas are NOT framed by dark non-NULL pixels as in the 64k monochromatic data.

While at some distance the Moon display looks quite pleasing, still when closing up, lots of bad mosaic transitions and color error patches exist. That's why I am still uncertain whether I shall ever consider publishing my results... Of course, if people want to experiment e.g. for making VTs ...let me know.

Here come some test shots that in my view look quite "natural". I feel overall this is the BEST available Moon texture at this time.












Let me know your opinion,

Fridger

[cartrite]

The screen shots look pretty good. A lot better than I thought was possible. I only put the 2 files together and I saw a lot of work was needed. I'm just going to keep the files as a base for LRO data which will available some day soon I hope.
cartrite

[Chuft-Captain]

I would be interested to see some closeups of the poles ... have you managed to get rid of most of the artefacts? That second image looks suspiciously like it might be one of the poles...

CC

[t00fri]

The screen shots look pretty good. A lot better than I thought was possible. I only put the 2 files together and I saw a lot of work was needed. I'm just going to keep the files as a base for LRO data which will available some day soon I hope.
cartrite

Steve,

indeed a lot of work is needed to comply with my usual standards.

But at least most black NULL pixel artefacts can be eliminated in < 1 minute via isis3-based cleaning (lowpass filter). The polar regime looks much better after replacing the NULL pixel and other very dark areas by a nice gray tone. Since the NULL pixels are not at exactly the same location in all three color bands, one has to again take recourse to the isis3 cleaning arsenal...

Fridger

[jogad]

Hi,

I feel overall this is the BEST available Moon texture at this time.

I agree with that.

You say there are artifacts but they are not so visible in the pictures. (Great shots).
But perhaps too beautiful?
Is it possible to have the original image in a format that is readable by your F-TexTools (or clues to do that) so we are able to have a personal point of view?

PS: Just a test? I hope no!

[t00fri]

I would be interested to see some closeups of the poles ... have you managed to get rid of most of the artefacts? That second image looks suspiciously like it might be one of the poles...

CC

CC,

Yes the north pole is visible in image 2 and N/S pole region also borderline in others.
That was actually the reason for showing image 2!

By replacing the black NULL pixels via isis3, the poles look MUCH more pleasing than
in the 64k raw texture.

Fridger

[t00fri]

Hi,

I feel overall this is the BEST available Moon texture at this time.

I agree with that.

You say there are artifacts but they are not so visible in the pictures. (Great shots).
But perhaps too beautiful?
Is it possible to have the original image in a format that is readable by your F-TexTools (or clues to do that) so we are able to have a personal point of view?

PS: Just a test? I hope no!

jogad,

as I wrote, all standard artefacts were eliminated with isis3 methodology. But this means that the many black areas were replaced by averages from surrounding pixels via e.g. an isis3 lowpass filter. This in turn means (cf. averages!) that from a somewhat more distant perspective, the display looks effectively FREE of artefacts, as is confirmed in my test shots. But when zooming in more, you will see plenty of localized color errors, sharp mosaic border lines, and the replacement areas from the previous null pixels.

Yes, it just takes a few minutes to generate a binary input for making VT tiles with my F-TexTools. As soon as I find some time, I'll do it and upload the resulting file. But remember, the upload will take several HOURS...

Fridger

[praesepe]

hey! You've cloned my previous post! Obviously I'm just joking

I'd say this is more than "just a test" the texture looks awesome even it's "only" a 16K one... This is definatively the best Moon texture so far. Maybe a low resolution version (1k, 2k) should replace the actual Moon map of the official distribution?

[t00fri]

hey! You've cloned my previous post! Obviously I'm just joking

I'd say this is more than "just a test" the texture looks awesome even it's "only" a 16K one... This is definatively the best Moon texture so far. Maybe a low resolution version (1k, 2k) should replace the actual Moon map of the official distribution?

Hi praesepe,

thanks for the flowers. Yet, it was also me who made the official moon texture in the distribution

But read my posts carefully. There are still quite a few artefacts on closer distance...

Cheers,
Fridger

[praesepe]

Fridger,

Ok, looking at the images you posted I can still see some artifacts. I had also a fast look at the .cub files at http://ser.sese.asu.edu/MOON/clem_color.html, all bands have them but at different locations, so I think that your results are more than reasonable according to the original data input... In a shorter scale the remaining artifacts wouldn't be so visible, say 1k or 2k which is the highest resolution map that comes with the standard distribution.

[t00fri]

Fridger,

Ok, looking at the images you posted I can still see some artifacts. I had also a fast look at the .cub files at http://ser.sese.asu.edu/MOON/clem_color.html, all bands have them but at different locations, so I think that your results are more than reasonable according to the original data input... In a shorter scale the remaining artifacts wouldn't be so visible, say 1k or 2k which is the highest resolution map that comes with the standard distribution.

Praesepe,

yes the NULL pixel areas are not at the same place. But we are ISIS3 experts aren't we? Hence there are three ways to try. I used 2 ways so far:

1) use 'bandtrim', which copies all NULL pixel areas (of each band) onto ALL bands. This takes a loooooooooong time, but works OK. Thereafter eliminate the NULL pixels with the standard lowpass setup like so:

lowpass from=<infile> to=<outfile> samples=130 lines=7 filter=outside null=yes hrs=no his=no lrs=no lis=no replacement=center

This way all local color artefacts will be gone, but you will have more patches with "average pixel" areas.

2) use 'explode' that will give you all 5 bands as separate files:

moon_5uvvis_500m.band0001.cub
moon_5uvvis_500m.band0002.cub
moon_5uvvis_500m.band0003.cub
moon_5uvvis_500m.band0004.cub
moon_5uvvis_500m.band0005.cub

Then eliminate the NULL pixel areas individually in each band with

lowpass from=<infile> to=<outfile> samples=130 lines=7 filter=outside null=yes hrs=no his=no lrs=no lis=no replacement=center

Then compose the result into a (r, g, b) PNG file with 'isis2std' and examine.

3) Ignore the different location of the NULL pixel areas and eliminate whatever you can with the lowpass. Do the rest by hand with "proven handicraft" technology

I am still examining which of the three approaches finally gives the most pleasing display!

I think (know) that the artefacts are visible at ALL resolutions (including 1k or 2k) once you zoom in more than corresponds to the "natural" distance at the given pixel resolution.


Fridger

[bdm]

But read my posts carefully. There are still quite a few artefacts on closer distance...

And if you zoom in reeeeealy close you might find artifacts like the Apollo 11 lander.

Even if the images do have artefacts, they are still very good, and bring the magnificent desolation of the moon to life.

[t00fri]

But read my posts carefully. There are still quite a few artefacts on closer distance...

And if you zoom in reeeeealy close you might find artifacts like the Apollo 11 lander.

But jokes aside, that's what people do and then come the complaints about a faint line here and a dark spot there ....

Even if the images do have artefacts, they are still very good, and bring the magnificent desolation of the moon to life.

I tend to agree with you! The really good aspect of these 500meter/pixel UVVIS textures is the sophisticated color. It really looks like the moon

Fridger

[t00fri]

Hi all,

unfortunately ISIS3 is still not available to Windows users. But perhaps it's of interest to some of you, if I illustrate a bit what the ISIS3 cleaning can do easily and what not.

Let me start will a little area of the colored 500 meter/pixel Moon from UVVIS/Clementine that includes a socalled NULL pixel domain. In my first example below, it is this black triangle shaped artefact in the image left of the blue bar:



The black color also indicates that this NULL pixel region is equally present in all three color bands that I used to form the colored moon image. Like this artefact, there will be many in our moon texture before cleaning. They are inserted by the software do indicate hardware failure or similar malfunctions. A NULL pixel can be a different number, depending on the used format. But anyway, ISIS3 has special features to deal with NULL pixels (and other special pixels) in the cleaning process.

So let us see explicitly what this simple ISIS3 command does to our black triangular artefact:

Code: Select all

lowpass from=moon_5uvvis_500m.raw.cub to=output.cub samples=130 lines=11 filter=outside null=yes hrs=no his=no lrs=no lis=no replacement=center


It is a lowpass filter that is configured to replace any NULL pixel by the pixel average of the surrounding filter rectancle (of size 130 x 11), called "boxcar"

The result you can see in the image to the right of the blue bar! The black triangular artefact has gone and clearly -- as seen from some distance -- the replacement will be hardly noted, due to the insert being an average of surrounding pixels! But some imperfection must of course remain since at any NULL pixel actual measurements were lacking!!

So on the good side, many such black NULL pixel artefacts will be instantaneously eliminated in the indicated manner with a single ISIS3 command.

Unfortunately the situation is not always that simple. In the next two little examples, you see colored artefacts which are also from NULL pixels that however are lacking or shifted on some color band(s). The composition of the bands then does NOT addd up to black. From the visible artefact color you can tell right away on which band the NULL pixels were.





As much as it looks, the best method seems manual cleaning as always
The GIMP has some neat tools for this, e.g the clone tool or the smudge tool.

Perhaps there is one other automatic method that I mentioned already in my previous post.
One can use an ISIS3 tool called 'bandtrim' which copies the NULL pixels to all bands. Clearly all colored artefacts can be automatically eliminated this way. So far I have used it for all 5 bands at once which resulted in an "artefact inflation".

But there is still hope: I suspect if I reduce from 5 colors to only three, the Iinflation" might well become much less severe...

Another hopeful aspect comes from applying the standard trick of converting RGB -> HSV (command rgb2hsv) and cleaning the artefacts in the 'Value' file.

We'll see...

Fridger

[John Van Vliet]

--- edit ---

[t00fri]

the above colored "stripes " are why i used a
red+green image to fix the blue
red+blue image to fix the green
and a green+blue image to fix the red image

the 5 layer cub ( i used the first 3 ) layer 1 =blue, layer 2=green, and layer 3 = red

There is a very easy solution within isis3:

1) 'explode' the 5 bands into individual .cubs
2) 'cubeit' bands 1,2,4 = (414, 750, 950) as a rgb cube just as in this official composition:
http://ser.sese.asu.edu/MOON/clem_color.html
3) 'bandtrim' the NULL pixels such that they appear on all three component textures.
4) replace the NULL pixel artefacts via a 'lowpass' by the 'boxcar' average.
5) 'isis2std' the result into PNG format

This takes about 15 minutes for the 21k Moon texture, altogether.

The result has NO multicolored remaining artefacts and is of equal quality as the official colored moon (500m/pixel) from here: http://ser.sese.asu.edu/MOON/clem_color.html

Fridger

[W0RLDBUILDER]

Will NASA please stop sending highlighters to the moon? It's making our maps all screwed up.