Celestia Forums

Hello:

In other threads I have documented my experience with a 32K normal map for earth.

When I created this normal map, I used a bump height of 100, which was suggested as the correct value for the GLOBE DEM data that was my raw data source. I assume this means that the resulting heights are accurate (i.e. not exaggerated.)

I have used this normal map on a blank grey surface texture, so I can study the normal map without interference from the main texture.

By comparing this view to the view of the Blue Marble texture, my observation is that the shadows on the Blue Marble texture are (subjectively) about ten times as pronounced as the shadows that are produced by the 32K normal map. Honestly, if I use the 32K normal map over the Blue Marble texture, it's hard to tell that a normal map is being used at all.

This leads me to one of two conclusions:

1) I did not specify a large enough bump height when I initially created the normal map. However, there are certain local features in my neck of the woods (Phoenix, Arizona) that look "right" in the gray orb / 32K normal map presentation. If the dynamic shadows were exagerrated from their current values, I'm pretty sure they wouldn't be realistic.

2) The shadows in the Blue Marble texture are not real, but have been exagerrated to emphasize depth.

Does anyone know for sure how the Blue Marble texture was created? Is there a high-res Earth texture available without any static shading?

Thanks for any info,

Tim Crews

Tim have you tried th Blue Marble texture available here
ftp://mitch.gsfc.nasa.gov/pub/stockli/bluemarble/
I'm not sure which zip it is but may be mod09a1Einterpolcylretouched3x21600x21600 and
mod09a1Winterpolcylretouched3x21600x21600
By the way where did you get the DEM data for your normal map?

My DEM data came from [url]http://www.ngdc.noaa.gov/seg/topo/gltiles.shtml
[/url]

See the thread "Experience with 32K Normal Maps" (or something like that) for more info.

Tim

jestr wrote:Tim have you tried th Blue Marble texture available here
ftp://mitch.gsfc.nasa.gov/pub/stockli/bluemarble/
I'm not sure which zip it is but may be mod09a1Einterpolcylretouched3x21600x21600 and
mod09a1Winterpolcylretouched3x21600x21600

Jestr:

Thank you very much for that link, it answered my question! From the readme file in that FTP directory, it appears that the BlueMarble texture that was the basis of Walton Comer's 32K virtual texture is the "3x shaded" version, I.E. it was shaded with a light source in the Northwest, and the bump map was exagerrated by a factor of 3.

The specific files that you named are advertised as "flat" maps, which hopefully means that they have no static shadows. I am downloading them now, and I will start working on making them into a virtual texture for use in combination with the high-res normal-map.

Tim

Tim,

1) Wow, I did not realize there was a shadowless version of the BlueMarble texture available. I wish I had known that back when I had the time to work on this. I would have definitely used that one instead of the 3x shadow version. That said, I cannot seem to access the ftp site for the unrendered version.
2) As far as the correct bump height goes the nm16 tool is really not adequate for accuracy. One reason is because that tool doesn't adjust for changes in latitude. Consequently at 60 degrees latitude all horizontal slopes are doubled, and this gets worse as the latitude increase. Secondly, 100 is way too large for earth. I don't recall the correct value but I think is was more like 0.1; regardless using the true earth shadow is very nearly pointless because you will not see any shadow features until you get down to the kilometer resolution scale. I would use something closer to 3x. PM me and I'll try to send you some sample code that corrects for the latitude and allows you to specify the bumpheight multiplier exactly.
3) I hope that Don. Edwards has access to the raw 32k files. I know for a fact that many of the ugly squarelike artifacts that show up through the Saudi empty quarter, as well as several other places, are due to artifacts in the topographic data. Unfortunately, all the 1k/pixel topographic data that I could find seemed to be sharing the same source material and have this artifact.

cheers,
Walton

Well if these new texture segments pan out I will changing over to them for the 32k texture I am making. A shadowless 32k texture would be a lot better than what we have had to work with. I will downloading them ASAP.

Don. Edwards

Walton I think there is a limit of 10 users at a time on the Nasa FTP site.I tried to connect earlier and got refused so I tried with a client FTP program and it said there were already 10 users-so keep trying,cheers Jestr

Yeah, it's probably the ten people following this thread!

Lot's of busy virtual texture builders tonight!

Tim

Well I have struck out. I downloaded the files I wanted, the un-shaded east & west along with the east & west clouds. The west clouds keeps coming back as an incomplete file. The others did download successfully but there sizes are being misrepresented in windows and I can't burn them to CD for transport. They all read as 2.000K files which is just wrong. I have opened them and then found that the .raw files are I believe 2gigs in size. I am not even sure I could open these in Photoshop with the amount of RAM I have. But than again I have heard of people opening big .raw files of Mars and working on them with less RAM than I have.
So as it stands I can't do anything with these files until someone can convert them to at least a 21600x21600 size .tiff or other format. Them is the breaks I guess.
So if someone can get the shadow-less images converted into a more friendly image type even if cut into large tiles I might be able to take the continental data from them and import it in place of the data I have from the older shaded data. So I and my 32k texture right now are the mercy of whoever can supply the goods.


BTW
If you have a look at the North-Pole.png on the site you will see that the oceans are colored black, as I have suspected this was the case in the originals. So I suspect the ocean coloring in the shadow-less images will in the end also be black. Which means we are going to really have to do some image manipulation. Or in my case it could make the transition to my texture even easier. Because if the oceans are black than selecting them by color and removing them should render a much cleaner transparency than the original BlueMarble textures did. Because of their re-coloring of the oceans left a great deal of black ocean data behind I have had to clean it up, which is why the texture is still being worked on.

As another possibility, I would be willing to turn over my ocean layer to whom ever wants to tackle this big project. I would include a detailed set instructions on how I was building and cleaning up my 32k texture and that way if someone wants to try it they can. Just give me the word and I can get things set up for whoever wants to try.

Don. Edwards

Hi all,

Don, glad to here you are attempting to add the shadowless texture. I'm sure the formatting issue can be solved so long as the actual data is not currupted. It shouldn't be too hard for someone to write an executable to fix the header information so that you can manipulate those files.

Tim, I looked through the original nm16 code and derived the necessary bumpheight factor for the true earth relief. Assuming your topographic data is in meters, then BumpHeight = Texture Width (pixels) / Earth Circumference (km) * 1 (km) / 1000 (m), else you need to use a different conversion factor so that your result is in pixels per unit of topography. For 64k texture, this works out to ~0.0016384 (pixel/m). Doubling the BumpHeight will double the slopes. But again this is not accurate for higher latitudes.

cheers,
Walton

I downloaded the eastern flat hemisphere earlier. The oceans are not black. I could'nt open it with photoshop or gimp so had to convert it to ppm format and then dice it up to view it. Not much else to say except that it is unshaded.

DBrady

Hi Don,

Correct me if I am wrong, but I thought you already had the shadowless texture; it is not new... I have them since the Blue Marble textures became available.
And did you tweak the settings of "scratch disks" under File/Preferences/Plug-ins & scratch disks in PhotoShop? Here you can set up swap space, not everything has to fit into RAM (although that would be much faster of course).

Hello:

My computer spent 5 hours today generating a 64K virtual texture (PNG format, 1K tiles -> 1.2G texture) based on the unshaded blue marble data.

I can confirm that the texture is indeed unshaded, and that the oceans are not black. They are _very_ blue.

The results look great with my (also just completed) 64K normal map. I am very pleased with the results. I no longer have any difficulty seeing the dynamic shading through the static shading, since the static shading no longer exists.

My spec map and night lights maps use 512-pixel tiles, which are currently incompatible with the 1K tiles of my surface and normal textures (because of a bug discussed in another forum.) So I had to disable them.

On my 1.2GHz Athlon, 1.25G of RAM, GeForce 4 Ti4200 with 64M RAM, I can get a 40fps frame rate while zoomed to the highest resolution.

I am thoroughly satisfied.

Once I am over my current burn-out, I would like to start another thread with instructional information and maybe some example scripts, so that everyone in the Celestia community (at least, whoever is willing to work with Cygwin) can generate these textures for themselves. It is probably not likely that we will ever see these textures hosted on the Internet.

As for myself, I don't even have a simple web page, which is why you don't see any screenshots in this post. [LATER EDIT: Actually, it turns out I do have some web space, and I have posted some screenshots in the "2MB of files inside!" thread in this same forum.]

Tim Crews

Let me clarify a few things. I was able to download the files to my brothers computer. His being my sole source of a broadband connection at this time. But his computer is far from mine. Speed wise it is just a little slower, an Athlon XP 2200 which was mine a few months ago. I love upgrading what can I say. But has no where near the RAM I do nor the Hard drive space. He has a single 60GIG drive that is nearly full, whiles have a 60, 80, and two 120s.
I was able to open the files and start the extraction but it was taking much to long and of course he didn't have the drive space. So I wanted to burn the files to CDR. Well that didn't work as Windows insists the files are only 2k in size rather than 220mb and 445mb in reality. So in the end I all got were two 2k useless files on a CDR. Now I can't e sure if the texture I am working on a shadow less or not. I have nothing to compare it to. I do see some very small shadows on some of the mountains here and there. Again without a point of reference its hard for me to tell. But if the BlueMarble East & West Topographical map that my 16k and 32k are based on the shadow less images you all are speaking of is in fact the same than I won't need to worry about it do I. So if someone who has downloaded them can take a look at the original 21600x21600 Topo maps and see if they are same than I could this all to rest. As I said my 32k is nearly shadow less as far as I can tell. So as I see it if we can get confirmation that the Topo maps and the shadow-less maps are in fact based on the same data than I am not going to sweat the details and I will go on and continue my 32k as is.
My offer of the ocean/water layer still stands for anyone that wants to try there hand at this. One thing to keep in mind is that the ocean layer can be resized up to say 64k and not loose any real detail as the water at this resolution still has no surface detail. I leave the decision up to thoughs that might want to try it. Just let me know and I can extract it, clean up a few things and zip it up.

Don. Edwards

wcomer wrote:2) As far as the correct bump height goes the nm16 tool is really not adequate for accuracy. One reason is because that tool doesn't adjust for changes in latitude. Consequently at 60 degrees latitude all horizontal slopes are doubled, and this gets worse as the latitude increase.

I stated the latitude issue poorly. Let me clarify, as the latitude increases the slopes become dulled unless adjusted. Using the nm16 tool, at 60 degrees the slope will be half the true value.

-Walton