Violet 0.1.3-rd (2k3 build for Wine users)

[rerdavies]

Could somebody who's running Wine give this one a try, please:

http://www.ottawajazzcollective.com/dow ... rd-2k3.zip

This is a visual studio 2003 build of Violet 0.1.2-rd. It's single threaded; and it will require vs 7.1 runtime DLLs. Cleaned up sources for visual studio 2003 will follow shortly, if I can get confirmation that this addresses the problem.

EDIT: Please note: This is not the preferred version of the violet binaries if you are using a Microsoft operating system. Please see the posting entitled

"Violet 0.1.3-rd Binaries + Sources Posted"

for more information.

[joegiampaoli]

Works and opens up perfectly fine without a single error, but I do see that Tone Mapping has changed quite a bit, like the power of values are almost 2X compared to what I get in blender exporter or last stable Violet I used.

[joegiampaoli]

Ok, here is a comparison with same igi and same ToneMap values.

Thanks for your work.

[rerdavies]

Ah. <grin>. It took me a second to catch on.

I *changed* units of the controls. They are now in STOPS (powers of 2), instead of in linear values. 1 stop = scale by 2, 2 stops = scale by 4, -1 stop is scale by 1/2, -2 stops is scale by 1/4. A convention that photographers are used to. And (I think), much more intuitive, especialy when you're trying to darken the image.

The post-scale slider, and the exposure slider on the linear tab use the same units. You'll also notice that the ticks on the histogram are in stops as well. Adding one to the Post-scale slider shifts the histogram one tick to the right; subtracting one moves one tick to the left.

A pre-scale value of 8 stops scales the input values by 256x! Why that big a range? Because it gives you roughly the same flexibility as linear tonemapping when using small values of "burn"; and it's improbable that the average value in a picture would be more than 8 stops away from where it should be. So the thinking goes anyway.

Your posts do raise the intersting question of why my setting of 4 (should be x16) is equivalent to the old setting of 8 (= x8). The default scaling of the image is supposed to place the average intensity of the image at 0.1, I think. Maybe I threw in an extra log, or an exp in there somewhere. I have to admit, I did notice on my test images that the default scaling for linear tonemapping was a bit darker than optimal. That being said, it's a difficult problem, usually solved, in cameras by multi-zone exposure meters, a healthy does of AI code, and a large japanese engineering department. But I'll check to see why I'm off by a power of two.

[jurasek]

Works fine here with Wine 0.9.36.
Tnx for your effort Rerdavies.



greetz,
jur