Vuescan preview inaccurate

[SJS]

Hi,

I am using Vuescan 8 with a Canon FS4000US scanner. The VS preview
seems coloured and does not match the raw file output.

Vuescan settings are :-
- image mode, 24 bit RGB
- exposure locked to 1
- color balance is none
- all brightness controls set to 1
- 24 bit raw file output with scan

The raw file has RGB 252, 245, 249
The preview is RGB 211, 223, 240

The top preview histogram shows the input values but the bottom
histogram shows the displayed values so I assume that it is the built-in
profile for the FS4000US that is causing the changes. Is this true ?

Can anybody tell me where I can get a profile (.ICC file) that will not
modify the scanner output at all ?

-- Steven

 

[Erik Krause]

Hello, SJS
you wrote...

The raw file has RGB 252, 245, 249
The preview is RGB 211, 223, 240

The top preview histogram shows the input values but the bottom
histogram shows the displayed values so I assume that it is the built-in
profile for the FS4000US that is causing the changes. Is this true ?

Yes and no. Raw files are in gamma 1.0 space. This is the gamma at
which the scanner delivers the data. Processed files are in a defined
color space for viewing. This is gamma 2.2 for PC color spaces (like
AdobeRGB or sRGB) or 1.8 for Mac ones (AppleRGB). "Built-in" is 2.2 I
believe. You can read at
http://www.poynton.com/notes/colour_and_gamma/GammaFAQ.html why.

The top histogram shows image data after conversion to your output
color space, the bottom histogram shows the data after vuescan color
balancing. Choose 'Color Balance None' for no change at all.

Can anybody tell me where I can get a profile (.ICC file) that will not
modify the scanner output at all ?

Use the raw files directly for that purpose. If you want IR-cleaned
and/or grain reduced raw files, choose 'Raw output with Save'. 'Raw
output with Scan' saves totally unprocessed data.

 

[SJS]

Yes and no. Raw files are in gamma 1.0 space. This is the gamma at
which the scanner delivers the data. Processed files are in a defined
color space for viewing.

My raw files are 24-bit RGB so I assume the 2.2 gamma is applied by the
scanner before Vuescan gets the data. However, a gamma adjustment
doesn't explain the difference between the original RGB and the preview
as Red was the highest but ends up the lowest value.

As the input slide is clear and exposure is locked on 1, I would expect
all 8-bit samples to be > 250 (255 - small amount to avoid clipping) and
the apparent scanner output is close to this. I can't see why Vuescan
is distorting the preview (to 211, 223, 240) and would like to stop
this. I believe I have disabled all colour adjustments in Vuescan and
the only variables left are the scanner profile and monitor profile.
Can I tell Vuescan to use an .ICC file that changes nothing in these
places ? If so, can anybody tell me where I can find such a file ?

http://www.poynton.com/notes/colour_and_gamma/GammaFAQ.html why.

Thanks for this link. It is very helpful

The top histogram shows image data after conversion to your output
color space, the bottom histogram shows the data after vuescan color
balancing. Choose 'Color Balance None' for no change at all.

I have 'color balance none' and brightnesses all set to 1 but still I
have the significant difference. Is the scanner profile applied between
the 2 histograms ?

Also, it seems that the 24-bit input from the scanner is affected by the
scanner profile before being written to the 24-bit raw file (output
during scan). Is this possible or do you think the differences I am
seeing are just due to scanner errors ? I am seeing a noise floor
(black slide) of 0 - 15 using one profile versus 0 - 22 using another.
The readings on the clear slide vary by plus or minus 4 (e.g. red is 239
to 247) but changing the profile has no effect.

-- Steven

 

[Erik Krause]

I believe I have disabled all colour adjustments in Vuescan and
the only variables left are the scanner profile and monitor profile.

Do you use Media Type Slide or Image? Slide only works with advanced
workflow, since in most slide images there is no clear film base
visible where the film base color can be determined. And even then you
will have to use Generic Slide to have no film specific adjustment.

Can I tell Vuescan to use an .ICC file that changes nothing in these
places?

If you scan from RAW there is different color adjustment depending on
the Input Mode setting. I believe 'Transparency' together with 'Built-
in Scanner color space' does no adjustment at all, since it's generic.
Ed, can you confirm?

Monitor and Printer color space do no alter file data.

If so, can anybody tell me where I can find such a file?

Since for vuescan you need a profile created by itself (it can not
interpret others) you only can do it yourself: go to
targets.coloraid.de and get a profile definition (.txt file). Use Scan
task: Make IT8 target to make a target image from that (preferably Raw)
definition.

Than use this image in a 'Profile Scanner' task. You can use the
qtProfileChecker to look at the profile whether it's really neutral.
It's part of the lcms profiler package:
http://www.littlecms.com/profilers.htm

I still don't see why this effort since you can use Raw files
directly...

 

[SJS]

Do you use Media Type Slide or Image? Slide only works with advanced
workflow, since in most slide images there is no clear film base
visible where the film base color can be determined. And even then you
will have to use Generic Slide to have no film specific adjustment.

Hi Erik,

Yes, I do use media type image. But last night the problem fixed itself
and my preview/scan view now matches the raw file. Yesterday I upgraded
from VS 8.0.2 to 8.0.3 and I changed some raw files from 48-bit to
24-bit by scaning from disk and creating TIFF files. After this my raw
scans/views are fine. I can recreate the problem by installing VS 8.0.2
to another directory and using it so maybe the calibrate function in
8.0.3 works better (I have an FS4000US scanner).

The odd thing is that changing the output color space changes the
preview/scan display whereas changing the monitor color space does not.
I now use 'device RGB' for output and 'Adobe RGB' for monitor color
space.

I do still have a problem with low value samples. Scanning a black
image in 24-bit mode produces 62% of samples at 0, and a total of 91% at
4 or less. The average sample value is 1. However, the bottom preview
histogram has a reduced peak at 0 (44% instead of 62%) and doesn't reach
a total of 91% until value 14. I'm not sure why this happens,
particularly the reduction of samples with value 0. This change
amplifies the apparent noise in dark areas.

My excessive interest in this area is because I am trying to ascertain
if the scanner is working properly.

Thanks,

Steven