Celestia Forums

I decided to start a new thread about this because the last 128k thread would not switch pages and ......... Well, I'm going to give this a go I think. Try to look past all the noise and other crap in these images and comment on if the resolution looks good enough. I'd like to know what others think.







If anything I can develop skills needed to work with these MODIS images so I can try to extract water body data above 60n latitude.
cartrite

I think such a thread would be perfect for CelestialMatters, where we started with all this (ISIS3 workshop, monster textures, normal maps,...) and where the home of my tools is, for example.... I will be over there, anyway.

Fridger

t00fri wrote:I think such a thread would be perfect for CelestialMatters, where we started with all this (ISIS3 workshop, monster textures, normal maps,...) and where the home of my tools is, for example.... I will be over there, anyway.

Fridger

I'm about to start one. Soon. I wrote a script that assembles srtm3 files in ISIS. I put together a signed 16 bit height map and had trouble matching the 64k map after using nm16. So I altered your NMS tool a bit and was able to use that to create a normalmap. It matches pretty good. I'm still testing and working out the bugs. Here are a few preliminaries. This is a 256k map. 90 meter resolution scaled down to power of 2. Normalmap is a peice of a 262144x131072 texture and the specnap is a peice of a 524288x262144 map.







cartrite

Pretty neat so far!

Here is a shot from a 128k surface texture, a 256k normalmap, and a 512k specmap.



cartrite

Amazing pics cartrite... last one looks a little noisy though?

It is Still a learning process and I got a ways to go. I haven't tried to clean it up yet. I was spending most of the time getting ISIS to reproject the MODIS images. Lots of work has to go into these yet.
cartrite

I think part of the problem here is that Celestia needs to reduce the level of detail level to avoid oversampling. Currently, Celestia picks the VT level based on the closest point and uses it globally. This has two bad side effects:
- More VT tiles than necessary are loaded
- There are visible aliasing artifacts on more distant tiles

I'm planning to do a rewrite of the sphere rendering code after 1.6.0. It will address not only these VT aliasing problems, but also some performance issues, depth sorting bugs, and the 'hole in the sky' bug.

--Chris

chris wrote:I think part of the problem here is that Celestia needs to reduce the mipmap level to avoid oversampling. Currently, Celestia picks the mipmap level based on the closest point and uses it globally. This has two bad side effects:
- More VT tiles than necessary are loaded
- There are visible aliasing artifacts on more distant tiles

--Chris

I was thinking something like this too. The tiles seem to be loaded now based on altitude. I often wondered what would it take to get them to load based on altitude and distance or distance alone. So when viewing the planets at angles like above, tiles in the distance would actually be lower level tiles.
cartrite

I took some screen shots running a 256k normalmap and 512k specmap of the western US and found these interesting.





cartrite

Nice, indeed!!

Fridger

cartrite wrote:I took some screen shots running a 256k normalmap and 512k specmap of the western US and found these interesting.

Mt-St-Helens-5.jpg

This is from my neighborhood The Columbia River looks great! Can you humor me and post a shot of Mt. Rainier?

chris wrote:

cartrite wrote:I took some screen shots running a 256k normalmap and 512k specmap of the western US and found these interesting.

Mt-St-Helens-5.jpg

This is from my neighborhood The Columbia River looks great! Can you humor me and post a shot of Mt. Rainier?

Steve,

but DON'T miss him out standing on top, smiling. I want to be able to read that Jolly Roger flag they were enrolling on top

Fridger

Unfortunately, at these high resolutions and low altitudes, high mountains tend to look like some large creature stepped on them when viewed from an angle. This is what motivated me to do Mars in 3d. But this looks pretty good even if Mt. St. Helens looks a little flat.



Creating this normalmap is not beyond anyones means in any way. For Mac and Linux users any how. Producing a 108000x54000 section of the planet that covers 45x22.5 degrees takes about an hour to process srtm3 files with ISIS. I'll post the latest script at the ISIS3 thread at CM.

Fridger, would you like a copy of the nmtiles that I modified to do partial normalmaps?

cartrite

Cartrite,

that last Mt Rainier shot is most beautiful...but your images are always on the dark side for me. So I need a call to GIMP to enjoy them, really.

Fridger, would you like a copy of the nmtiles that I modified to do partial normalmaps?

Yes, please! I'll have a look at the code. It might well be a little tricky

Fridger

cartrite wrote:Unfortunately, at these high resolutions and low altitudes, high mountains tend to look like some large creature stepped on them when viewed from an angle. This is what motivated me to do Mars in 3d. But this looks pretty good even if Mt. St. Helens looks a little flat.

Thanks--very impressive work, though the mountains do indeed look a little squished. I'm pretty sure that breathing would be a bit easier on the summit of this version of Rainier But, for 1.7.0 I plan to rewrite the sphere renderer, and we should finally have real terrain rendering on a global scale.

--Chris

chris wrote:But, for 1.7.0 I plan to rewrite the sphere renderer, and we should finally have real terrain rendering on a global scale.

--Chris

Indeed...that's really missing (besides good atmospheres )!

Fridger

chris wrote:

cartrite wrote:Unfortunately, at these high resolutions and low altitudes, high mountains tend to look like some large creature stepped on them when viewed from an angle. This is what motivated me to do Mars in 3d. But this looks pretty good even if Mt. St. Helens looks a little flat.

Thanks--very impressive work, though the mountains do indeed look a little squished. I'm pretty sure that breathing would be a bit easier on the summit of this version of Rainier But, for 1.7.0 I plan to rewrite the sphere renderer, and we should finally have real terrain rendering on a global scale.

--Chris

I was wondering how this can be accomplished. Are you thinking of using height maps for the elevations?
I am also a novice in this area so this may seem like a dumb question but is it possible to use a height map instead of a cmod model?
The model list the x and y coordinates in the file. Couldn't they be derived from the pixel position in the file. Also, would simplecylindrical maps be necessary? Line 1 of a simple cylindrical map could only contain 1 pixel. Line 2 contain 8, etc.
cartrite

cartrite wrote:

chris wrote:

cartrite wrote:Unfortunately, at these high resolutions and low altitudes, high mountains tend to look like some large creature stepped on them when viewed from an angle. This is what motivated me to do Mars in 3d. But this looks pretty good even if Mt. St. Helens looks a little flat.

Thanks--very impressive work, though the mountains do indeed look a little squished. I'm pretty sure that breathing would be a bit easier on the summit of this version of Rainier But, for 1.7.0 I plan to rewrite the sphere renderer, and we should finally have real terrain rendering on a global scale.

--Chris

I was wondering how this can be accomplished. Are you thinking of using height maps for the elevations?
I am also a novice in this area so this may seem like a dumb question but is it possible to use a height map instead of a cmod model?
The model list the x and y coordinates in the file. Couldn't they be derived from the pixel position in the file. Also, would simplecylindrical maps be necessary? Line 1 of a simple cylindrical map could only contain 1 pixel. Line 2 contain 8, etc.

Yes, the idea would be to use height maps instead of cmods. Cmods are more general than height maps, and thus it's much more complicated to implement level of detail and culling algorithms with them. Height maps would be probably be stored as tile hierarchies, much like virtual textures. Normal maps would still be useful for shading of the height maps. Anyhow, these are just some initial thoughts--there's a lot more research and discussion of this topic is necessary.

--Chris

Here are a couple screenshots from a new cloud map I'm doing. It's from day 243 from this year. August 30. I'm downloading the images from the MODIS Rapid Response System at http://rapidfire.sci.gsfc.nasa.gov/realtime/2008243/ and the LAADS WEB at http://ladsweb.nascom.nasa.gov/data/search.html I'm using a tool to process the images called HDFLook.

I'm having a little trouble getting it to be less transparent though. They don't look right when you can see right though them. Only the first shot really shows this transparency though. I got tired of looking at that other map from 2001. There are also some other images I posted at CM.







cartrite