
www.icegraphics.com The main forum for the ICEGraphics software

View previous topic :: View next topic 
Author 
Message 
vreijs
Joined: 19 Jun 2004 Posts: 21

Posted: Tue Jun 22, 2004 11:30 am Post subject: accuracy of computations and linear color space 


Hello CIE,
I was wondering what is the accuracy you do your calculations, is this 8 or 16 bits for each color channel? The more the better of course.
Furthermore, I understand that in general one can do calculation in the color space only if one is in a linear color space. Do you also go from the gamma related space to a linear space, then do the multi dimensional arithmatic (color warping) and then back with a gamma to nonlineair space).
With what (device dependant) gamma do you go to lineair space (what value)? In my own optimaztions it is part of the process to determine the device dependant gamma and after the manipulation of the colors in lineair space, I go back to a gamma of 2.2 (sRGB space).
Thanks for your feedback.
All the best,
Victor 

Back to top 


ICE Graphics Site Admin
Joined: 31 Mar 2003 Posts: 431

Posted: Sun Jul 04, 2004 3:18 pm Post subject: Re: accuracy of computations and linear color space 


vreijs wrote:  I was wondering what is the accuracy you do your calculations, is this 8 or 16 bits for each color channel? The more the better of course. 
Source image, of course, 8 bit per channel. Due to BMP or JPEG limitations. For internal calculations use "long double" type, it's mean 80 bits per channel.
vreijs wrote:  Furthermore, I understand that in general one can do calculation in the color space only if one is in a linear color space. Do you also go from the gamma related space to a linear space, then do the multi dimensional arithmatic (color warping) and then back with a gamma to nonlineair space). 
ICE Color can work with any color space, linear or nonlinear. So, ICE Color use only direct gamma correction. I mean, it's apply gamma correction to image only once.
vreijs wrote:  With what (device dependant) gamma do you go to lineair space (what value)? In my own optimaztions it is part of the process to determine the device dependant gamma and after the manipulation of the colors in lineair space, I go back to a gamma of 2.2 (sRGB space). 
ICE Color know nothing about (device dependant) gamma. Ant this is very good solution. Because, (device dependant) gamma not always is known. 

Back to top 


vreijs
Joined: 19 Jun 2004 Posts: 21

Posted: Sun Jul 04, 2004 9:22 pm Post subject: Re: accuracy of computations and linear color space 


[quote="ICE Graphics"]
ICE Color know nothing about (device dependant) gamma. Ant this is very good solution. Because, (device dependant) gamma not always is known.[/quote]
Thanks for your responce, at least I understand you thought about them and that is for me the most important. Thanks.
All the best,
Victor 

Back to top 


vreijs
Joined: 19 Jun 2004 Posts: 21

Posted: Sun Jul 04, 2004 9:39 pm Post subject: Re: accuracy of computations and linear color space 


[quote="ICE Graphics"][quote="vreijs"]I was wondering what is the accuracy you do your calculations, is this 8 or 16 bits for each color channel? The more the better of course.[/quote]
Source image, of course, 8 bit per channel. Due to BMP or JPEG limitations. For internal calculations use "long double" type, it's mean 80 bits per channel.
[/quote]
I see what the reason is!!! I had originally a tif file and made a jpg file for working in ICE COlor, but that was not a good choice of mine, I should have converted it to a bmp file. If I do that, the results is much better. Sorry about this!!! Thanks for your assurance of your internal accuracy!
All the best,
Victor 

Back to top 


vreijs
Joined: 19 Jun 2004 Posts: 21

Posted: Fri Jul 09, 2004 7:01 pm Post subject: Re: accuracy of computations and linear color space 


[quote="ICE Graphics"][quote="vreijs"]Furthermore, I understand that in general one can do calculation in the color space only if one is in a linear color space. Do you also go from the gamma related space to a linear space, then do the multi dimensional arithmatic (color warping) and then back with a gamma to nonlineair space).[/quote]
ICE Color can work with any color space, linear or nonlinear. So, ICE Color use only direct gamma correction. I mean, it's apply gamma correction to image only once.
[/quote]
Perhaps I used the wrong wording. I am [b]not[/b] talking about the gamma you can determine/input in the Level adjust window.
I am talking in general; A file (like jpg , tif or bmp) has normally the gamma already in it for the color space it was made in (so in sRGB space, a gamma was used of appr. 2.2 by default).
So a file in jpg/tif/bmp is not encoded in linear color space, but has a gamma by default. If one wants to manipulate a picture, one normally first gets the encoding out of the nonlinear/gamma environment (by taking the gamma=0.4545) over it, then manipulate it and then back to the gamma of the color space (for sRGB gamma=2.2).
Beside these inherent gammas (of .4545 and 2.2) one can of course still apply another gamma (which is the one you are talking about in the 'Level adjust' window).
It is possible that your method of adjusting/rendering the file has totally no problem with this inherent gamma of 2.2, but it could be that in a linear space it performs better. Did you test this out?
Thanks for your feedback.
All the best,
Victor 

Back to top 


ICE Graphics Site Admin
Joined: 31 Mar 2003 Posts: 431

Posted: Sat Jul 10, 2004 7:34 am Post subject: Re: accuracy of computations and linear color space 


vreijs wrote:  So a file in jpg/tif/bmp is not encoded in linear color space, but has a gamma by default. If one wants to manipulate a picture, one normally first gets the encoding out of the nonlinear/gamma environment (by taking the gamma=0.4545) over it, then manipulate it and then back to the gamma of the color space (for sRGB gamma=2.2).
Beside these inherent gammas (of .4545 and 2.2) one can of course still apply another gamma (which is the one you are talking about in the 'Level adjust' window).
It is possible that your method of adjusting/rendering the file has totally no problem with this inherent gamma of 2.2, but it could be that in a linear space it performs better. Did you test this out? 
ICE Color do not apply gammas (of .4545 and 2.2), like you wrote. But anyway, you can applya it manually.
1. Just apply gamma to file 0.4545, save it.
2. Open in ICE Color. Change this is file.
3. And after color correction in ICE Color apply gamma again  2.2.
Or even you can combine steps 2 and 3. Because ICE Color apply gamma correction after color space warping. 

Back to top 


vreijs
Joined: 19 Jun 2004 Posts: 21

Posted: Sat Jul 10, 2004 7:50 am Post subject: Re: accuracy of computations and linear color space 


[quote="ICE Graphics"]
ICE Color do not apply gammas (of .4545 and 2.2), like you wrote. But anyway, you can applya it manually.
1. Just apply gamma to file 0.4545, save it.
2. Open in ICE Color. Change this is file.
3. And after color correction in ICE Color apply gamma again  2.2.
Or even you can combine steps 2 and 3. Because ICE Color apply gamma correction after color space warping.[/quote]
Oke, I have tested this and if I first do a 0.4545 gamma then rendering and then 2.2 gamma, the average error (deltaE) is somewhat larger then when using only rendering (not really a significant difference to be honest).
So it seems that ICE Color has no real problem to work in the nonlinear space I would think.
All the best,
Victor 

Back to top 




You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum

Powered by phpBB © 2001, 2005 phpBB Group
