Celestia Forums

Is there a (undocumented) option to change the brightness
of the nightside textures , possibly in the solarsys.ssc file ?

Most NightSide textures are almost "binary" with no graylevels between
0 and 255 . I like the nightside textures , but I think they are way to bright
compared with the "daylight" textures


Ren?©

mine was made as an artistic rendering and not to be correct you will nead to dim the map in gimp or photo shop

rra wrote:Is there a (undocumented) option to change the brightness
of the nightside textures , possibly in the solarsys.ssc file ?

Most NightSide textures are almost "binary" with no graylevels between
0 and 255 . I like the nightside textures , but I think they are way to bright
compared with the "daylight" textures


Ren?©

Are you sure that your "ambient light" is close to zero!? Check the { (less) ,} (more) keys.

It's really easy to adjust these textures in brightness with GIMP. You can select particular colors and adjust these separately. A 1 minute affair, really. It takes much longer to write posts about it

I always do my nightlight textures (32k VT's) from scratch. Go to the nightlights site (easy to find with Google) and download the big raw images. Then make a small model texture to try out the best brightness/contrast settings. Then you may decide to leave the lights either white (mono texture) or you use some bright yellowish color. You may become much more fancy of course, by adding some "structure" into your extended city lights with GIMP. A simple technique with a big effect (just using a textured layer underneath).



Bye Fridger

this is part of levle 2 vt of mine
i got it from hear
http://dmsp.ngdc.noaa.gov/html/download.html
ti is a 16 bit geay scale image in .dat form . Imagemagick ( 16bit not the normal 8 bit for windows ) will convert it to a 16 bit .png that can be worked on in gimp or tiles can be worked on with nip/vip from the The London National Gallery
http://www.vips.ecs.soton.ac.uk/index.php?title=VIPS

I produced a 32k virtual texture of Earth night-lights with full tile optimization, but I?€™m not sure it deserves distribution, so I need some expertise to validate my approach. This is the description.
The map is based on the ?€?Radiance Calibrated Lights: 1996-1997?€™ from NGDC: [url]http://www.ngdc.noaa.gov/dmsp/data/rad_cal/rad_cal.tar[/url]. Accordingly to documentation, ?€?Radiance = DN^(3/2) * 10^(-10) watts/cm^2/sr?€™, where DN is the binary value in the pixel. On the other hand, for the standard PC screen, radiance = brightness ^ gamma, where gamma = 2.2. So I took brightness = 255 * (DN/255) ^ (3/(2*2.2)). So, the radiance of a given point on the computer's screen should be proportional to the measured intensity of light sources on the Earth. Accidentally, this produced a luminosity level, which appears to me quite acceptable aesthetically and in comparison with other light sources in Celestia.
Then I applied the BMNG water mask to refine shorelines and inland water bodies.
Finally, I derived the coloration from the NASA?€™s photo of New York City:

There are my attempts to reproduce this photo in Celestia:
New York in midnight (without clouds):

New York in the morning:

New York in the evening:

Additionally:
Over Europe (without clouds, with Fridger Shrempp's ?€?earth-Gazetteer-2.ssc?€™ add-on):

Over Japan (without clouds):

Certainly, there are some problems. I see questionable very bright spots somewhere in Pakistan and in Gobi desert. But such are the data!
What does the community think?

Unexpected bright spots often are fires of some kind: oil wells or forests.

Since today there is some interest to nightlight textures, I try to revive this tread. Can someone say something about the validity of my approach or is someone interested in the subject now?

Well, I hate being ignored too, so...

The celestia outcome doesn't quite seem to have the impact of New York city's lights, but...

If you 'pre-emphasise' the virtual textures for monitor gamma and then compare it to the NASA image, which is not 'pre-emphasised', you'll never get a match! I recommend you compare them before pre-emphasis, and then pre-emphasise the VTs.

If you want to be really pedantic about monitor gamma, then you'll be dismayed to learn that PC monitors use a different gamma to Macs (see W3C: PNG and Gamma: Cross Platform Woes (http://www.w3.org/Graphics/PNG/platform.html) for further gory details). You'll also see that PNGs have a means to convey the correct adjustment to be used for correct monitor gamma. It's in the PNG specs somewhere. See sections 2.7, 12.7 and 13 of PNG Spec. 1.2 (go to Portable Network Graphics (PNG) Specification, version 1.2, http://www.libpng.org/pub/png/spec/1.2/).

Even then, it turns out that that's not foolproof: [url=http://hsivonen.iki.fi/png-gamma/]The Sad Story of PNG Gamma ?€?Correction?€

Hello, Spif.

Thank you very much for your post. Now this is my answer.
I chose this particular data set mostly because, unlike most other light maps, city centers here are not saturated, and some details can be distinguished. But the file description claims, that binary value in each pixel is a (dimensional) function of radiance: ?€?Radiance = DN^(3/2) * 10^(-10) watts/cm^2/sr?€™! So, the direct use of this file is impossible anyway, this is rather data table, than an image. I described my method of transformation. Obviously, the parameter ?€?Gamma?€™ may by different (any user can adjust it according personal needs, after all), but Gamma=2.2 is the most common (and Macs are not), as far as I understand. The photo of New York shows that, for my mind, the result is not so bad. If you can propose something better (transforming the data, testing the result by comparing with the NASA photo) ?€“ please, give an advice.
As for color, I doubt that worldwide data on this issue do exist. The use of this photo as source for coloration of the texture is, at least, not at all arbitrary. Again, please, advice me something better.

With best regards, Robert

Sorry not to reply sooner, but I got the feeling I'm being latched onto as a 'master' and I'm not!

RVS wrote:The photo of New York shows that, for my mind, the result is not so bad.

I agree the comparison is not bad at all, especially given that a lot of textures used and added to Celestia are not checked for any calibration or such at all. So yes, you got lucky for your work with a good comparison. However, I think the astronaut's photo shows more contrast and dynamic range: New York centre is much brighter, if you look carefully.

That's why I suggest comparing before pre-emphasis, yet your idea is sensible. The comparison would fall down in a rigorous sense because we don't know what the characteristics of the astronaut's digital camera are. Being likely CCD based, I'd expect linear. Thus you should compare your image to it before pre-emphasis.

Sorry to sound negative, but in my mind, my best idea is this: it's an excellent piece of luck to get this data in 'raw' radiance form. In an 'ideal' Celestia, it would take this data and create the night lights accordingly. This is similar to another idea I had: that stars shouldn't be rendered directly with textures of their surfaces already coloured etc., but should have grey-scale images that represent black-body temperatures across their surface. Celestia would convert each pixel into the correct RGB setting. However, I tend not to raise such ideas because I cannot code them, and - well - you see that perturbing the 'to do' list of someone else for such an esoteric idea is perhaps impolite.

RVS wrote:As for color, I doubt that worldwide data on this issue do exist. The use of this photo as source for coloration of the texture is, at least, not at all arbitrary. Again, please, advice me something better.

Ah, what can I say? I too doubt that there is a single repository of such colour data and I note that these images of Earth at night have uniform colour: probably because they are colourised black and white images.

I know from flying over the UK at night, towns and cities there glow with a stark orange colour. Over Germany, it's a more creamy colour.

Here's a (silly) idea though! Use the Celestia forum's worldwide distribution of observers and set up a query so that people can say what kind of streetlamps are used in their country.

Spiff.

Hello, Spif. Thank you again, your last post is very constructive.
On the subject?€¦
Ideal Celestia, running on ideal computer with ideal graphics board, ideal monitor, ideal operating system and ideal picture formats would be certainly ideal. But in reality everything has its limitations. Monitors do not create picture with the radiance, proportional to the brightness of the picture and with the known fixed coefficient of this proportionality. When we want to show a picture on the computer screen, we need to somehow map the entire brightness range on the narrow gap, that monitors can represent. There is an infinite number of ways to do it, and the outcome can be appreciated only subjectively. However, as practice witness, it may be quite satisfactory. My goal was to create as far as possible realistic texture with as little arbitrariness as possible. But some is inenvitable, and some compromise is required. The absolute fit with the photo is unreachable because of multiple reasons, of course, I posted the link only to allow the estimation of the quality of the texture. I tested the transformation with Gamma=1 too, the result was ?€¦ sad.
As for gathering of information about the outdoor lightening, this is a theme for a good scientific work! I am afraid, the user interrogation will give very incomplete coverage. Of course, a lot of spatial, air, and ground-based photos of city lights can be found with the ease in the Net, even on the same NASA site. The spectral patterns of light sources can be found as well, but this is obviously overkill. My chose of this photo is motivated mostly by esthetics. Again, I think, that the color of my texture is more realistic, than that of most others.
What about the fact, that my texture is resolution-optimized, and monochrome (i.e. black, in this case) tiles are not included? About 5.5Mb in zip archive, plus about 8Kb for 8k version.
In brief, I am ready to distribute it, but only if I obtain the quality approval from the community, and if I will see, that someone needs it.
With best regards, Robert

RVS wrote:Ideal Celestia, running on ideal computer with ideal graphics board, ideal monitor, ideal operating system and ideal picture formats would be certainly ideal.

True! Though I meant ideal in the way that Celestia code would take actual calibrated data and convert it to correct appearance, rather than take a given texture and slap it on a body, before we worry about hardware...

RVS wrote:In brief, I am ready to distribute it, but only if I obtain the quality approval from the community, and if I will see, that someone needs it.

Heh! I think we're not that formal about it! If you look at the Celestia Motherlode http://www.celestiamotherlode.net, you'll see that people put many things on there with varying degrees of standard. Some, such as Don Edwards have a pride that means they only realise things they are happy with, too. So you see, it's up to how you feel about your work.

I'm not sure anyone can ever organise more than 1% of the whole Celestia community to give quality approval! We swap, take or leave things, come and go... .

Now I notice it, the specular map seems to make 'holes' in the texture that look like moths have eaen it. Is there really a need to refine shorelines if the raw data was truly observed radiance, as in even the sea must be lit up by nearby cities...

Personally, I do agree the use of orange is much of an improvement!

Spiff.

Spaceman Spiff wrote:
Heh! I think we're not that formal about it! If you look at the Celestia Motherlode .

I would like to see, if some unacceptable flaw in my methodology will emerge (so far, it wasn?€™t). On the other hand, if this product will pass ?€?community expertise?€™, it will encourage users to download it.

Spaceman Spiff wrote:Now I notice it, the specular map seems to make 'holes' in the texture that look like moths have eaen it. Is there really a need to refine shorelines if the raw data was truly observed radiance, as in even the sea must be lit up by nearby cities...

This is a good question. The brightness of light diminishes very quickly (proportionally to the inverted square of distance from the source), so from above the light source itself and it?€™s nearest neighborhood can be visible. The water, in turn, is usually very dark and absorbs light very good. The pixel size of 32k texture in Celestia is more than 1 km. Reflections on the water surface are possible as well, but it can produce only ?€?duplicates?€™ of shoreline lights below water level in the best. Light sources on the water (in the areas of intensive navigation, fishing or off-shore industrial activity) are possible, but it undoubtedly should be much darker, than city lights. On the same photo of New York shorelines, Hudson River and other water streams are well distinguishable. My conclusion is that water on the texture should be dark. But, for some reason, this data file itself doesn?€™t show any traces of rivers, and offshore areas are frequently polluted by light. I can attribute this only to inaccuracy. This is why I applied the water mask to the lightmap. BMNG watermask is imperfect too, this is well known to Celestia users, but the better is unavailable.

With best regards, Robert

Hello to all.
Since nobody expressed interest in obtaining my night map (although I still don?€™t see a way to improve it), I don?€™t offer it moreover. However, here is a small set of photos of night-lights, which (as well as the discussion above) may be useful for future creators. Of course, more shots may be found with ease:


Northeastern Seaboard of the U.S.:

Seoul, South Korea:

Los Angeles, California:

London, England:

New York, the same photo once more again:

Finland, Russia, Estonia and Latvia:

Chicago:

With best regards, Robert

Robert,

Just because people don't say anything doesn't mean they aren't interested. Most people are just too shy to speak up in public. The only way to measure how much people are interested is to look at the Web server's logs to see how often the file is downloaded.

I think the real photo of Chicago is a good indicator of just how far current night maps have to go. (no offense Robert - I'm talking about all night maps I've seen in general)

Chicago:


Only a couple of features in this image are what in my opinion make it so realistic:
- Not all lights are the same colour - subtle colour differences are noticeable (radiance alone is not enough)
- Sharp grid features - all night maps I've seen are too blurry

As RVS noted, even in a 32k texture the resolution is too low - 1km per pixel is just not enough when trying to portray city streets. There's probably no remote sensing data of high enough resolution either, but as Fridger suggested it's possible to hand-draw street lights and make your own - but of course it can no longer be considered realistic.

the best data i found was from the defense department ( declassified ) data
at 42x21 k in gif format
so hand drawing is going to have to be done
as to coloring i did ad some variation for the different kinds of lights
and more work is still needed

dirkpitt wrote:Not all lights are the same colour - subtle colour differences are noticeable (radiance alone is not enough)

Surely. But what can we do? ?€“ There is no global color photo coverage. Once again: my approach is, at least, not entirely arbitrary.

dirkpitt wrote:even in a 32k texture the resolution is too low - 1km per pixel is just not enough when trying to portray city streets. There's probably no remote sensing data of high enough resolution either, but as Fridger suggested it's possible to hand-draw street lights and make your own - but of course it can no longer be considered realistic.

john Van Vliet wrote:so hand drawing is going to have to be done

The typical dimension of a block in the city is of order of several hundred meters (~600 m in the case of photo of Chicago in consideration). Over-imposed grid will exaggerate sizes many times and can?€™t be considered realistic. Individual streets just can?€™t be resolved ?€“ as well as in the case of daytime surface maps. Celestia is not Google Earth.

selden wrote:Robert,

Just because people don't say anything doesn't mean they aren't interested. Most people are just too shy to speak up in public. The only way to measure how much people are interested is to look at the Web server's logs to see how often the file is downloaded.

Oh-oh! If somebody will agree to write methodology description, installation manual and upload files to some public server (ML is acceptable), achieving so the status of co-author , I am ready to e-mail them to (s)he. I don?€™t want to do all this for a number of reasons (starting with, but not limiting by, my imperfect English).
With best regards, Robert

RVS wrote:The typical dimension of a block in the city is of order of several hundred meters (~600 m in the case of photo of Chicago in consideration). Over-imposed grid will exaggerate sizes many times and can?€™t be considered realistic. Individual streets just can?€™t be resolved ?€“ as well as in the case of daytime surface maps. Celestia is not Google Earth.

I wouldn't give up yet - you can make out streets in the VT Level 11 and 12 closeups on the motherlode. For example, look at Andrea's Venice addon (daytime only) - canals are certainly resolved, and gondolas might be visible with some squinting.

Here are renderings of locations from the space photos on the previous page in Celestia with my night map, accompanied by URLs and the content of 'favorites.cel' file for camera positions and orientations. You can compare real photos with views with whatever night map you are using and judge what is more realistic yourselves.

Chicago, London, Los Angeles, New York, Seoul:

dirkpitt wrote:I wouldn't give up yet - you can make out streets in the VT Level 11 and 12 closeups on the motherlode.

Yes, local close-ups are possible, but this is a different matter. We are speaking about global maps of entire planet, aren?€™t we?
With best regards, Robert