Image Editor - Lack of Smooth Gradient

[Barry Thomas]

I've rendered about 40 images for product presentation over the past few days via 'batch' rendering directly in nXt and with the nXt Farm with all out put as .nXtImage files so that I could further manipulate the images as need before inserting into various documents.  As I'm performing post production on the images, I realize there is a hard line gradient in all of the images displayed in the Image Editor. All images saved as JPG from the Image Editor also have these hard line shadows.

I've attached screen captures of the output. The first image is a capture of the render window and represents how JPG output from nXt appears as well, smooth gradient on the floor. The second image is a capture of the Image Editor screen and also represents how the JPG out put looks when saved from the Image Editor.

I got the same results with both builds 283 and 284 Image Editor.

[Roy Hirshkowitz]

When you get a chance-- try saving from the Image Editor to png.  This will help me determine if this is a jpg compression problem or some other Image Editor problem.

[Barry Thomas]

Thanks for looking Roy. Same result -see attached

[Roy Hirshkowitz]

I just read your message a little more carefully-- it does sound like the nXtImage is bad.  

This isn't happening here-- until I can make it happen I wouldn't even know where to start.  See if you can come up with a simple procedure to reproduce the problem.  If you can, stop and upload all of the necessary files and explain the procedure.

[Barry Thomas]

you can grab the file here: https://secure.logmein.com/f?00_lPUwd1h8H9TcNMi9Vza.3drWgZ4yaxNwgrR...

It's about a 7MB download. This is the DWG with related files in an ACAD eTransmit folder.

The unorthodox approach I took to lighting this model is most likely the source of the problem when it hits the Image Editor, and the reason you'll need to see the original file. I tried to replicate this with a smaller model, but was unable to get the same lighting effect. It's an old school quasi light dome set-up that I've been using for years and often resort back to when I do not have an HDR providing the needed effect.

So... render the model "VIEW 1" (I did at 3300x3300, but resolution doesn't seem to matter. I tried it at lessor resolutions and got the same result)

The number of passes don't seem matter either. My original was a 20 passes. I've tested this at 40+ and as few as 6 and have gotten the same result.

save the output as a JPG or PNG and you will see the image background is smooth, colors on the model are correct, wood grain represented properly.

save the output as an nXtImage.

open in the Image Editor - you will see the "stepped" shadowing in the background. The colors of the wood grain will appear over saturated. The image actually looks like it is being processed with 256 colors as opposed to 16 million.

Thanks again for looking Roy.

[Roy Hirshkowitz]

Ah-- if you've got an nXtImage file hand I can take a look at that and may be able to tell right away-- without running your model.  It's possible that the actual luminance values using your method are either very low or very high.  The nXtImage file does compress the original floating point data using a similar method to .exr files.  This is normally not a problem if the lighting in the drawing is in a "normal" range.  However, very low values can result in a precision loss and very high values are clipped.

[Roy Hirshkowitz]

Never mind-- it was obvious once I ran the model.  As suspected-- the lighting values are way too low.  You'll need to use a lighting scheme that brings that up to something reasonable if you want to use the .nXtImage format.  If you move your cursor around the image and see a lot of zeros for values in non-black parts of the image, you're in trouble.

[Barry Thomas]

That did it, I increased the light source to 900,000 lumens. It was at just 12,000. Even at 900k, the values were at .001. I increased the lighting to 2 million lumens and get a reading of .012 on the floor. Even so, at 900K, the image made a trip through the Image Editor just fine.

I know this has been asked and answered before, but what is a good average to shoot for? and can you give to me, again, the formula for deciphering the lighting values at the bottom of the screen.

Thanks again, for your help. Glad it was my bad, and not your code.

BW

[Roy Hirshkowitz]

The Luminance units are in cd/m^2 and have a huge range.

An average gray surface in direct sunlight will measure something like 15,000.

The same surface on a cloudy day might be 1000.

Indoors, in a bright, daylit interior (out of direct light) and you might get a value like 500-- under artificial light at night, perhaps 50.

Moonlight on a forest floor might be 0.001.

You can find typical values if you poke around online-- there are some here.