Vuescan Color Balance voes

Hello,

<br><p>

 

I'm having trouble getting consistent (never mind good) color from

negatives (but sometimes also slides) from Vuescan and a Minolta 5400

scanner. The root of the problem seems to be Vuescan's color

correction algorithm whose results apparently depend on what is

included in the desired crop. The following examples, all screenshots

taken from the preview screen of Vuescan, illustrate the problem:

<a

href="http://tachyon.uwaterloo.ca/~plangfelder/VuescanProblem/WhiteBalance:Neutral.jpg">This

preview screenshot</a> has White Balance set to "Neutral" and the

color is alright, nothing that could not be fixed later. Note that

part of the dark area on the right is cropped out (the crop area is

delineated by the thin line). On the other hand,<a href

="http://tachyon.uwaterloo.ca/~plangfelder/VuescanProblem/WhiteBalance:Neutral-2.jpg">

This preview screenshot</a> has the exact same processing parameters

applied except that most of the frame is included. Now the color is

nothing short of useless.

 

<p>I get similar results when I select White Balance "None": <a

href="http://tachyon.uwaterloo.ca/~plangfelder/VuescanProblem/WhiteBalance:None.jpg">This

screenshot</a> has part of the dark area cropped out and would be

usable, whereas <a

href="http://tachyon.uwaterloo.ca/~plangfelder/VuescanProblem/WhiteBalance:None-2.jpg">this

one</a>, differing from the previous one only in including more of the

dark area on the right, is useless. I have encountered this behaviour

a number of times by now and for certain situations it simply renders

Vuescan useless (I scan under Linux, so vuescan is my only option).

 

<p>Can anyone please explain (or point me to an explanation) how

Vuescan calculates color correction/color balance? Is there any way to

disable the white balance correction and have it apply only infrared

cleaning and film color conversion? (the film color conversion tables

cannot depend on crop, can they?) What causes the color to go so

horribly wrong?

 

<p>Thanks for all replies,

I've just started scanning color negative film with my 5400. I tried:

 

1. a 100% Vuescan workflow, first outputting 64bit Vuescan Raw Files

 

Problem with above, I could not get decent cleaning quality, with any of the cleaning filter settings.

 

2. a 50/50 Vuescan/Minolta Scan Utility workflow, doing the initial scan with Minolta Scan Utility, outputting 16 bit linear, and then using this as a Vuescan Raw File.

 

Problem with above, I could not get a decent color balance. Perhaps it can be done. I have experimented a lot with color slide scans, and am quite familiar with the tricks. Tried advanced workflow, etc.

 

3. a Minolta Scan Utility to Photoshop workflow, as follows:

 

Scan with Minolta Scan Utility as:

 

As a Color Negative

 

16 bit linear tiff output

 

Auto Exposure on

 

Auto Focus on

 

No crop or rotation

 

ICE/GD on

 

Then, in Photoshop,:

 

Clean up dust and scratches (missed by ICE)

 

Crop to exclude frame (No preset size, just as close as practical to edge, with crop tool. Don't do any slight rotations prior to crop, to compensate for slight skew, it may degrade image)

 

Rotate portrait images 90 degrees

 

Save (to preserve unaltered original), then "save as", and in the copy:

 

Image|Mode|Assign Profile:

 

MINOLTA DIMAGE SCAN ELITE 5400 (POSI LINEAR)

 

Image|Mode|Convert to Profile:

 

Adobe RGB 1988 (with intent "relative colorimetric")

 

 

Image|Adjustments|Invert (hotkey: <ctrl> i) (reverse the histogram)

 

Apply levels with:

 

.01/.03 wp/bp per channel

 

with snap nuetral midtones <BAM! Spot-on color, at this point

 

Arbitrary: Apply unsharp mask:

 

300%

 

.3 pixel radius

 

0 threshold

 

Convert to 8 bit and save

Regarding the applying cleaning to unalter image, you can do that if you set the Vuescan Raw File to save "at save", as opposed to the more usual save "at scan". Warning, be very careful not have a custom ICC Profile, just use "built in". There is a Vuescan BUG, contrary to the helpfile, the profile IS being applied to the raw data. I recently posted a rant on the subject.

 

Something you might look into, if you continue to use Vuescan, the Crop|Buffer setting affects how much of the preview window is considered in exposure calculation. The default is 15%. This means a 15% strip around the edge is ignored. I believe the distance from edge to center constitutes 100%.

Woops, just noticed you say your system is Linux. Assume that means Minolta Scan Utility is not an option?

Peter, not sure if this can help, but it is what I do. I have been using VueScan for several years now and find it fantastic.

 

Under the 'Color' tab, set your colour Balance to 'White Balance.' This is best general setting. I notice the examples you pointed to appeared to be some sort of artificial light, maybe a camp fire? Under those conditions, try 'Tungsten', or experiment with other settings.

 

Sometimes, setting your 'White Point' to zero, rather than the default of '1' will pick up a lot more highlight detail. The scan will then be flat, but all detail will be retained. Afterwards, adjust in PS to get the tonal balance that pleases you.

 

I have found that much 'fiddling' beyond what I have described can make it worse. Remember, VueScan is for 'scanning.' PS is for 'adjusting.'

Peter -

 

Use the batch scan workflow described in the help file. Essentially, you scan a blank frame at the beginning of the roll to lock the exposure time and mask color. Then scan a 'typical' frame, and lock the color balance. When you do this, your options in the color tab change and are limited - vuescan no longer tries to change things on the fly based on the crop. Instead, it uses the locked settings. Then you can scan the entire roll with the same settings. Of course, this assumes that all of the frames were shot in similar circumstances and are not badly exposed.

 

For the example you posted, you could use the cropped region of the frame you're happy with to lock the color settings. Then everything else on the roll should be OK.

Why not just scan full frame and crop later?

Chad, thanks a lot for the suggestion; that's precisely what I was looking for.