Optimizing landscape resolution

[tomjwilson]

I have been making animations of spacecraft flying over Mars terrain, using Image Mapping to place high resolution photos taken by real Mars orbiters onto a rolling-hills Object I made in Modeler. My problem is even the highest res photos of Mars terrain rendered onto the smallest possible Object appear too pixelated. I have tried moving the terrain away from the camera which makes the terrain look smoother, but the terrain Object has to be much larger when it's farther from the camera. I can't seem to balance keeping the terrain Object as small as possible, and at the same time make the pixels look smooth enough for a sharp animation. Should I be modeling the terrain 100% and forget Image Mapping the big, hi-res Mars photos onto the modeled surface or what? Or, is there some how to get a window on scaling the grid, terrain, spacecraft that I don't understand?

[richpr]

Is surface smoothing on for the surface?

And how big is your image map of Mars?

[tomjwilson]

To answer your questions, I check SurfEd/Smooth, and use Mars terrain photos sized between 600 and 6000 Megs. My LightWave grid is 1m. I have bought LW books and run tutorials about landscaping, but I don't understand what the scales of the spacecraft flying and the terrain below should be. Which direction should I start from, first size the terrain & then the spacecraft, or reversed? What approach gets you the best results? Or maybe you know of a book that covers this?

[Dave]

could you give the pixel dimensions of your images?

thanks,
Dave

[tomjwilson]

Dave: The jpg I used last is only 680 kb, with 1114w x 570h pixels.

Thank you!

[gjjackson]

I wonder if it's a matter of scaling of X,Y,Z.

[lwgal]

I've noticed that zoom can sometimes have the effect that you're talking about with pixelization....even using a 12000X6000 pixel image, if you turn up the zoom a lot, a great looking image will have problems zooming in on it.

The other problem is the size of your image compared to the size/scale of your object. If you are using (for example and I always try to build at a realistic scale for normal sized type objects and terrains) a 600X600 image on a 6km piece of terrain, you will only get 1 pixel for every 10 meters or roughly 30 feet. If you try to fly down close to that terrain, the image map will look awful. The image maps of Mars would do great probably 1km into the air above a piece of regular sized terrain if the images are big enough in size (maybe even higher), but they are not detailed enough for a low flyby.

You might try adding some procedurals over the image map to break up the surface some and hide the pixelization better....or perhaps size up the map appropriately (it might end up to be monstrous) and run some image processing algorithms on the image to smooth it out some (perhaps soften) and then try it again that way with the smoothing out from a larger image.

Failing that, you might have to paint an image map of your own, cloning details from the original to a much larger canvas and building up a much larger and more detailed image...that usually works pretty well for me in dealing with terrains.

[starbase1]

I have done stuff with Mars surfaces before - but never tried to use an image map because, as others have said, the detail really isn't available at the resolutions you are attempting.

But you might be aboe to do it if you pick a big enough feature - for example, Olympus Mons - that is so huge that an image map could work very nicley.

But you will probably find you get much better results using gradient based textres - for example, by making steeper slopes look rougher, and flat areas lighter, (the idea being this is where dist collects). I have also experimented with applying a small amount of specular to the steeper slopes, with quite good results.

Soft srata like colouring works well as the bottom layer ion the colour channel.

Or how about using a soft version of the image map? That way there would be an underlying accurate colour, without pixels shoping, and you can then overlay the gradients.

Cheers,
Nick