Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[cameni]

Since each refinement level is in separate wad files, we can easily have the demo mode using only 76m data (that's how I created the 76/30 screens), and have the 38m refinement as a downloadable option.

Even 152/30 dataset is quite usable, some types of games might use that to get even smaller data size.

Point is, the real resolution of 30m SRTM data is lower than 30m because of the smoothing, it seems to be somewhere much closer to 76m and therefore it's not as big improvement as I had hoped. I focused on 90m vs 30m comparison in the videos and screenshots, but most of the added detail is visible in 76/30 data. It's a big enhancement compared to 76/90 data, as the 90m sources were heavily filtered, washing out the detail at resampled 76m. If they were "true" 90m data, it would look like what you saw in 76/30 screenshot.

[HiFlyer]

You know, I've always had a nagging feeling that the Middle earth demo was somehow more detailed and naturalistic than regular Outerra, but I didn't understand how that could be. Now I wonder it its because their elevation data was "pure" with less artificial smoothing built in. Peaks and sharp edges always seemed more clear, there.

It's all very interesting, and even more interesting how you will choose to at least partially "correct" the data, if thats possible.

[ZeosPantera]

The Middle earth data is also more extreme. There are quite a few cliff faces that make real world cliff faces pee their pants.

[HiFlyer]

The Middle earth data is also more extreme. There are quite a few cliff faces that make real world cliff faces pee their pants.

Yup! I think the main thing is that there are far more "sharp" areas like ridgelines, and less of a "smoothed" feeling. Its a subtle difference, but you can feel it. I think a lot of us waiting on the new dataset were probably hoping for sharper definition like in this picture of Middle earth ridges. I hope there is a good middle ground.

[KW71]

Have you downloaded the data from earthexplorer.usgs.gov?

I don't understand why they would bother in scanning a higher resolution, and then smooth it.


- Could be in the site you downloaded the data (or other site), another untouched-non-smoothed version?

- Would some subtle sharpening filter, help to enhance de data?

- Converting the data from 16 bits to 8 (or increasing the contrast) would result in sharper steps and thus, in sharper forms?


If the answer is "no"  to previous questions, I guess is clear you should go for the 76/30.

[KW71]

For example, just a small adjust in Gimp (render in Blender, exactly same settings for both):
NOTE: It is just a picture from internet, not data used in OT.


ORIGINAL / CONTRAST ENHANCED:



ORIGINAL:




ENHANCED:




EDIT: ENHANCED LESS CONTRAST

[HiFlyer]

The question is, does that make it more realistic? Or just artificially more irregular?

[KW71]

The question is, does that make it more realistic? Or just artificially more irregular?

Is just a bit of contrast in the image. The adjust could be even smaller (see edited images), and, of course, this renders lack the fractal detail.

[HiFlyer]

I think the less contrast images look better, but other peoples milage may vary. Outerra straddles a fine dividing line. I'm just worried that things don't become too "fractally" and cross the line into over-detail.

[2eyed]

Well, I'm a bit dissapointed with the outcome from socalled 30m data. Had higher hopes regarding this. Sharpness of cliffs and ridgelines and more detailed erosive patterns would help to make terrain look more realistic.
Mabe there is a way to refine your procedural/fractal engine without extra data input (at the expense of world fidelity).

[PytonPago]

Well, I'm a bit dissapointed with the outcome from socalled 30m data. Had higher hopes regarding this. Sharpness of cliffs and ridgelines and more detailed erosive patterns would help to make terrain look more realistic.
Mabe there is a way to refine your procedural/fractal engine without extra data input (at the expense of world fidelity).

 ... well, ill say, in mountains, 1 m resolution would still be not enough (seriously, 10 meter wide 90-depressions cant be taken out from a 30-m resolution) .... i doubt that even 10 meter would be any totally significant change - you just wont have them in such detail to walk the path like in real tourist-track matching them as perfectly. But, personally, im quite ok whyte how they are, even if those hill-tops aren't as "flash-backs" invoking as i have some of them in memory. General shapes are pretty good compared to the real world ....

[KW71]

... well, ill say, in mountains, 1 m resolution would still be not enough (seriously, 10 meter wide 90-depressions cant be taken out from a 30-m resolution) .... i doubt that even 10 meter would be any totally significant change - you just wont have them in such detail to walk the path like in real tourist-track matching them as perfectly. But, personally, im quite ok with how they are, even if those hill-tops aren't as "flash-backs" invoking as i have some of them in memory. General shapes are pretty good compared to the real world ....

+1

[josem75]

Well, I'm a bit dissapointed with the outcome from socalled 30m data. Had higher hopes regarding this. Sharpness of cliffs and ridgelines and more detailed erosive patterns would help to make terrain look more realistic.
Mabe there is a way to refine your procedural/fractal engine without extra data input (at the expense of world fidelity).

 ... well, ill say, in mountains, 1 m resolution would still be not enough (seriously, 10 meter wide 90-depressions cant be taken out from a 30-m resolution) .... i doubt that even 10 meter would be any totally significant change - you just wont have them in such detail to walk the path like in real tourist-track matching them as perfectly. But, personally, im quite ok with how they are, even if those hill-tops aren't as "flash-backs" invoking as i have some of them in memory. General shapes are pretty good compared to the real world ....

I remember myself, testing 5m resolution in my region, in FSX almost one decade ago.
I would say, the fidelity of the shapes was nice. Some small peaks was there while they was not in the original 90m or even 25m. So result was closer to real life.
But i agree. The result is great comparing the video posted here.  Its a big Step in how terrain look. Coverall Avoid a lack of details in the flat places. And make mountains better, even visible with the shadows generated. Now when you look the terrain in the side of 30m, is much more realistic.

I dont know what should be necesary to get more real shapes compared to real life. Maybe there is some way to get more raw data from the 30m SRTM.  OR some filter can be added.

In 2020 we will be playing Outerra in 5m. Or maybe it will have a way to add certain places (as in Fsx) with really accurate data.
And maybe in the near future from now, with the tools for retouch the terrain and layers (rocks, etc), we can create accurate places. making a extrusion here and there, or a smooth. Changing some rock patterns, etc.
I still need to see the 30m in the places i visually know. I still think it will be more accurate data compared to 90M. we have much more points in the middle.

Still vote for have a experimental 30m version to explore and refine data.

KW71 solution is not looking bad. I preffer his last image compared to the original.

[KW71]

Cameni, are you using the full 16 bits for height resolution?

I guess eight bits is enough for current database


Mariana trench: 10,809 m  +  Everest: 8,848m   = 19,657 m          / 256 = 76.78m


But 30m database would benefit from more bits.

[cameni]

You are confusing vertical resolution with the horizontal one. Vertical is in meters and world would look pretty ugly if we quantized it into 8 bits.
Horizontal spacing is around 30m, which gets resampled into ~38m OT dataset.