This can't be that hard... IT8 Profiles

steveo_mcg

steveo_mcg

Messages
6,320
Name
Steven
Edit My Images
Yes
How the **** do you get it square in the scanner. I've been at it an hour and I can not get the stupid thing to line up with the guide on screen, the skew doesn't seem to help. Off to do something else for 10 minutes before I need to buy a new scanner...
 
Make it easier on yourself, just scan the target and then use LProf to create the profile:

http://lprof.sourceforge.net/

All you need to do then is scan the target, save it and input it into Lprof where you select the corners of the target and it plots on the image where it is reading from, then create the profile which you assign to scanned images. Its a lot better than using Vuescan's built in profiling (which is very simplistic and requires the target to be near dead straight). (Read here for an easy guide on how to use it http://lprof.sourceforge.net/help/lprof-help.html)

When scanning the target, make sure that you are not overexposing any of the patches (use the 'pixel colours' settings on the colour tab), and then lock down the exposure, colour settings etc (as obviously changing any of them makes the profile incorrect); try adjusting the exposure time as far as it will go before you start to get clipping and adjust down/up until you reach the maximum you can. Save the scan as a TIFF (as the jpeg compression can screw up the profiling sometimes) with a wide gamut colour profile like Adobe Wide Gamut (make sure 'tiff profile' is selected in the 'output tab').

Hope this makes it a bit simpler, if you need any more help then I'm try to be of assistance.

Sam
 
If you have issues with LProf (it wouldn't run for me) try CoCa which is also free

Just get your scanned image in photoshop or similar then straighten it if required then crop to the marks and just put it into CoCa.
 
s162216 said:
Make it easier on yourself, just scan the target and then use LProf to create the profile:

http://lprof.sourceforge.net/

All you need to do then is scan the target, save it and input it into Lprof where you select the corners of the target and it plots on the image where it is reading from, then create the profile which you assign to scanned images. Its a lot better than using Vuescan's built in profiling (which is very simplistic and requires the target to be near dead straight). (Read here for an easy guide on how to use it http://lprof.sourceforge.net/help/lprof-help.html)

When scanning the target, make sure that you are not overexposing any of the patches (use the 'pixel colours' settings on the colour tab), and then lock down the exposure, colour settings etc (as obviously changing any of them makes the profile incorrect); try adjusting the exposure time as far as it will go before you start to get clipping and adjust down/up until you reach the maximum you can. Save the scan as a TIFF (as the jpeg compression can screw up the profiling sometimes) with a wide gamut colour profile like Adobe Wide Gamut (make sure 'tiff profile' is selected in the 'output tab').

Hope this makes it a bit simpler, if you need any more help then I'm try to be of assistance.

Sam
Click to expand...
Thank you! I'll try that when I get home from work. With all the knowledge on the web it still comes down to knowing who / what to ask...
 
When you install Lprof make sure to check the readme as it lists the things you need to have installed for it to run. Its powered by LittleCMS so obviously you need to install that: I just tried running it and it worked (I had previously installed LittleCMS).

You need to assign the profile to each scanned image in photoshop or similar. I only have elements which gives only a choice of sRGB or Adobe RGB, so I open the scanned images in the freeware RawTherapee as that has a profile assign feature, save it in there and then open in PS elements (make sure to select a wide gamut output colour space in RawTherapee if you use it though).

Sam

-- Sent from my Palm Pre3 using Forums
 
So would I open it in RawTerapee, save it with a wide gamut then contiune processing it in LR as normal? Its either that or GIMP and I'm quicker in LR...
 
abbandon said:
If you have issues with LProf (it wouldn't run for me) try CoCa which is also free

Just get your scanned image in photoshop or similar then straighten it if required then crop to the marks and just put it into CoCa.
Click to expand...

Just been reading some more on that? I take it out put from CoCa is more akin to scanner input profile much like ViewScan produces?
 
steveo_mcg said:
So would I open it in RawTerapee, save it with a wide gamut then contiune processing it in LR as normal? Its either that or GIMP and I'm quicker in LR...
Click to expand...

Yes, if I had the full Photoshop then I would do it all in there, but for what I do it would be overkill (and bloody expensive!).

steveo_mcg said:
Just been reading some more on that? I take it out put from CoCa is more akin to scanner input profile much like ViewScan produces?
Click to expand...

All of these programs do the same thing (broadly speaking) which is creating a scanner input profile, they just do it in different fashions. Vuescan's IT8 profiling is very simplistic in comparison to all the other options though.
 
s162216 said:
Yes, if I had the full Photoshop then I would do it all in there, but for what I do it would be overkill (and bloody expensive!).



All of these programs do the same thing (broadly speaking) which is creating a scanner input profile, they just do it in different fashions. Vuescan's IT8 profiling is very simplistic in comparison to all the other options though.
Click to expand...


Thanks Sam. Once I get this working I'm going to have a lot scanning to do!
 
@steveo_mcg Got it working then? Remember that in Photoshop etc you will have to set the levels for the scanned images as obviously the profile doesn't know what should represent the maximum white level etc. Hope that your scans come out better now.

Check your profile with the 'profile inspector' option in Lprof and use the dE tab take a look to see how well your profile does for each colour patch, <3 is seen as being optimal and practically no different from the reference. I would expect some of the darker/more saturated ones to show higher values, but when creating a new profile (you can just use the same scan as before) you can adjust the profiling by altering the settings in the 'profile parameters' box (one thing I would do is select the '33 - huge' resolution option as it gives a slightly better resolution profile). I found that lowering the average deviation setting a bit balanced it out for my profiles as previously the dE value for the darker tones was much higher.

Sam
 
Last edited:
s162216 said:
@steveo_mcg Got it working then? Remember that in Photoshop etc you will have to set the levels for the scanned images as obviously the profile doesn't know what should represent the maximum white level etc. Hope that your scans come out better now.
Click to expand...

Yes, much better. There is still a magenta cast on a few of them but it might be the stock (it was ood) and they were shot through an old folder so exposure might be off. I can probably photoshop the cast out, but its there in the slide so thats not the fault of the scan. That really is the best thing about slides, the original is a near perfect reference guide.


Check your profile with the 'profile inspector' option in Lprof and use the dE tab take a look to see how well your profile does for each colour patch, <3 is seen as being optimal and practically no different from the reference. I would expect some of the darker/more saturated ones to show higher values, but when creating a new profile (you can just use the same scan as before) you can adjust the profiling by altering the settings in the 'profile parameters' box (one thing I would do is select the '33 - huge' resolution option as it gives a slightly better resolution profile). I found that lowering the average deviation setting a bit balanced it out for my profiles as previously the dE value for the darker tones was much higher.

Sam
Click to expand...

I'll need to re-read that whilst I'm in front of the program as that has left me scratching my head. Broadly though do you make a profile for different exposure levels? I've done this one at a brightness of 2.4 in ViewScan (lowest that had no clipping) so would I do one at say 3 to apply to a dark frame to get more shadow detail.
 
Just to be certain, in Vuescan did you set the exposure by using the 'lock exposure' option under the input tab (not the brightness setting in the 'colour' tab)? You don't need to make different profiles for different exposure levels as thats not the idea of IT8 (imagine the scanner as a fixed exposure slide copier - it doesn't correct for the source image, it tries to make an exact copy).

You can however help improve shadow detail in dark transparencies by using the 'multi-exposure' option in Vuescan when scanning, and also by manually combining exposures using Photomatix (the 2 images exposure fusion option is always free if you download the trial version) before applying the profile. Theres an excellent guide about how to do that here: http://www.jingai.com/scanningguide/slide exposure blending.html (about half way down the page).

Sam
 
Yeah locked the exposure, you gave me that tip last time I tried this.

I think I must have got the wrong end of the stick with the last post, I'll reread it when I'm in front of my pc.
 
Chaz Photos said:
I am interested in some of the comments about profiling scanners.
I profile all my gear camera screen and printer but not found a way of doing the scanner.
Can anyone tell me more
Click to expand...

You need to buy an IT8.7 target (be aware that this only works for transparencies or prints, you can't do it for colour negatives due to the differences between the colour masks with each film) such as the ones that Wolf Faust sells:

http://www.targets.coloraid.de/

Obviously if your scanning transparencies then get a transparency target, and a print one if your scanning prints. You don't need to get a target for every film type as one will usually work quite well for all films, so just get one for the film you use the most. Each target consists of a series of colour patches, and after scanning the target you use a program like Lprof to compare the scanned patches to the reference values provided by the target manufacturer (who will have done densitometry analysis for each patch). From this a colour profile is created which you can apply to scanned images and will help to get scans look more like the originals (it is very difficult to get an exact scan, but the profiling is an immense help).

If you have any more questions then I'll try to answer them,

Sam
 
You must log in or register to reply here.
Back
Top