DSE5400 - Give me a proof to calibrate it

[gilou]

Hi there,

 

Yesterday I've received my OPTIX XR Calibrator, and have calibrated

my CRT (Viewsonic P95f).

Before buying a calibrator i was not really convinced by the

effectiveness of such device, but i'm obliged to say that my photos

looks really better on the screen now.

 

My question now is relative to the scanner profile (DSE5400). I've

already asked question about that, but i would like to see from a

real user the differences between a profiled Scanner and a non

profiled scanner.

 

Could someone scan a slide (Sensia/provia) and then show me the

results with the Scanner profile and without the scanner profile

applied to the file (without any other processing under PS or any

other tool)?

 

I would like to have a proof, and see if there is also a big gap

between a DSE5400 Calibrated and a non Calibrated Minolta 5400.

 

thanks for your support. Brgds -Gilles.

[mendel_leisk]

Gilles, I'll post some examples when I get home tonight. In the interm, some comments.

 

If you recollect, I described my workflow in your earlier post regarding profiling. How I output 16 bit linear thru Minolta Scan Utility, then scan-from-disk these, with Vuescan.

 

In the Vuescan HTML help file, Ed Hamrick suggests that once you have profiled your scanner, and have pointed Vuescan to the ICC file, if you want to see the results as the ICC file has changed the color balance, use "Neutral" setting in the color tab.

 

I've re-scanned-from-disk thus with a few rolls, using the "Neutral" setting and the ICC profile I've produced. The results ARE consistantly "neutral".

 

Toggling between my previous efforts, with "White Balance" setting, and the new, sometimes they are ALMOST indistinguishable, sometimes there is quite a marked variation. This is due in part, to the effect "white balance" was having, more so on some pictures than others. "White balance" was partially correcting my uncalibrated ouput, but "messing" with color balance as well, sometimes with undesirable results. If I tried "neutral", prior to calibrating the scanner, the results where anything but neutral.

 

The net result after calibration is consistant clean up of the reddish cast I was getting, and, coupled with "neutral" setting, images that stay much truer to the slides (in my case).

 

I have some color negative film in my camera, and intend to photograph the large target I also purchased, see how that works. A workflow for this is also described in the Vuescan HTML.

[mendel_leisk]

All 3 of the following are produced with Vuescan from the same raw file. The first uses the ICC scanner calibration file, The second uses "built in" with Neutral color balance, the third "built in" with white balance.<div></div>

[mendel_leisk]

Built in with neutral<div></div>

[mendel_leisk]

Built in with white balance<div></div>

[gilou]

thanks a lot Mendel,

 

As usuall you help me a lot on the "perfect scanning workflow" path.

 

It is clear that the calibration of your scanner, removes the red/magenta colorcast. Now I know what I need to do..... spend 40 EUROS and buy these IT8 targets.

 

For your info, I've tested your workflow (Scan with the Minolta software and then uses VUESCAN), and my results are getting better. and I supose that I will get better ones as soon as my scanner will be calibrated.

 

Another time thanks a lot.

[gilou]

I've tested a funny thing ..... AUTO LEVEL .....

And the result is that the 3 photos are so close together now, that it is difficult to distinguish the calibrated one from the others, however, the picture1 has a more natural green grass color than the others !

 

To me the 3 Auto level pictures are great to me .... so where is the point to calibrate the scanner ! (Joke)

 

Awayting your feedback .

 

Try it ..... AUTO LEVEL ...

[mendel_leisk]

Assume by Auto Level you mean Photoshop's process to clip the red, green, and blue channels by equal percents. That would help to remove casts in a lot of cases I would think.

 

Perhaps the object with calibrating scanner is to attempt to replicate on-screen the original colors of the slide. And using auto levels may clear up cast, but shift the color balance away from the original slide, in the process.

 

Just making this up as I go, Gilles. I've had my targets for less than a week, and really don't know that much on the subject. The guy who responded to your last thread on the subject (falling of a log comment) could perhaps contribute. I'd be interested to know his workflow.

[mendel_leisk]

Well, I made my LATEST icc profile completely through Vuescan, following the instructions in the Vuescan Help File.

 

Before, I made my scans through the Minolta Scan Utility, outputting 16bit linear. This time, I scanned through Vuescan, and worked directly with the preview image. I set preview res to the lowest availabe preset.

 

Note, you can overide the file names/path for the it8 and icc files by entering something else (more descriptive) in the color tab. The fields are:

 

Color|Scanner ICC Profile

 

Color|Scanner IT8 Data

 

This is all explained in the help file, but so tersely I didn't understand it at first.

 

In concurrence with other Elite 5400 user reports, I find my slide scans are a bit skewed, about 1/2 degree counter clockwise. So, I just kept remmoving the holder and rotating the slide clockwise until I got it as aligned.

 

I took my time lining up the grid, zoomed in a bit.

 

Following this workflow, my latest icc profile produces these changes:

 

1. Slightly warmer (I think towards neutral, scrutinizing the original slides)

 

2. A little brighter overall.

 

3. Slightly higher contrast (as evidenced by histogram in PS)

 

4. Recovered detail from highlight and shadow. This last is very nice, really improves the quality.

 

Here's the difference:<div></div>

[mendel_leisk]

I meant to mention also, by "Auto Level", did you mean Vuescan's Auto Level?

 

It will really "level" out the color balance, but I think at the expense of being true to the original slide colors. There can be a real shift, and results often look posterized, especially with some of my tungsten light shots.

[gilou]

hi Mendel,

 

thanks for your comments and support.

 

To answer your question, I'was speaking about the PS auto-level (or any other application) not the one from Vuescan.

I do agree that the auto-level changes the overall colors of the slide ! I've done that many times and there is a huge gap between my slide on the light table and the result i can see on the screen.

 

Good job !

Your picture looks great now. There is no color cast, the grass is green ans the xhite part of the jumper is white !

the contrasts are better ...

 

Conclusion : I'm going to order today this IT8 Target !

 

gilles.

[mendel_leisk]

As I said, my best/latest ICC profile was made exclusively through Vuescan. I did two, one with the Fuji Provia target, and one with the Kodak Ektachrome target. Even though I suspect the film is Ektachrome, I got better results with the Fuji Target. It seems somewhat hit or miss.

 

I'd suggest to get the "all you need" package (from Wolf Faust), and then experiment with all of the targets.

 

This latest icc profile, from the Fuji Target, really does seem to match the physical slides. And that's my object.

 

If you're using Vuescan, you will see that when you change ICC profiles, it changes the BEFORE histogram (the top one).

 

Cheers!

[gilou]

Hi Mendeil,

That could come from the fact that your slides are not so young ..... and the color has shifted with time. This could explain part of the problem.

Give a try with fresh developped slides.

Just my 2 cents - Cheers.

[pawel_bartuzi]

Reading your conversation I thought I'll add my two cents. I also use Minolta DSE 5400 (first version) with Vuescan and I am in no way a professional scanner operator however I spent some time profiling my DSE 5400 a couple of weeks ago. To evaluate color fidelity I was comparing my slides put on a lightbox with the same slides scanned and displayed on a monitor calibrated and profiled with hardware calibrator.

 

First, when using the scanner with built-in profile of Vuescan and color balance set to neutral I noticed my scanned slides had consistently red/magenta cast - very annoying as these were from my winter holidays with lots of snow that looked pink. :-)

 

After reading for a while on the subject I bought an "Agfa IT8 Color Reference" target set and began to profile the scanner using Vuescan. The results using Vuescan produced ICC profile and color balance set to neutral were better, however the cast was not gone completely and when comparing the slides on a lightbox with the scans I still could see a difference.

 

Together with my IT8 sets I got for free a long discontinued profiling software - Agfa Colortune 3.02 (now Agfa is selling only its professional version costing several thousand euros). As the software was over 5 years old and I was using the latest and newest version of Vuescan I didn't even install Colortune. However, seeing still not perfect results with Vuescan I decided to give Colortune a try.

 

I scanned an IT8 target with Vuescan (with color balance set to none or neutral and ICC setting set to default "built-in" profile) and then generated scanner ICC profile using Colortune. I tried to set this profile as a scanner ICC profile in Vuescan, but it didn't work so I rescanned my slides using exactly the same Vuescan settings I used when scanning the target and then attached my Colortune generated ICC profile to the TIFF files produced by Vuescan. And... it made all the difference! My slides on a lightbox looked now _exactly_ the same as the scans on a monitor! On some slides only minor brightness and contrast adjustments were neccessary - the color was spot on.

 

What I learned afterwards is that Vuescan produces very small, simplificated ICC profiles of inferior quality - so called matrix profiles (mine was only about 500 bytes long!). Agfa's Colortune instead produced a "complete" ICC profile with size over 600 kB containing detailed color lookup tables (CLUTs). These tables are used by photo processing software to interpret and display the colors outputted by Vuescan and they describe the scanner behaviour much better than matrix profiles produced by Vuescan. Of course, the big Colortune profile doesn't have to be attached to the TIFFs indefinitely - after using the profile as an input profile I'm free to convert my TIFFs to whatever color space I wish and forget about the scanner profile.

 

So, if you want maximum color fidelity for your scans you should give some serious scanner profiling software a try. Agfa's Colortune is not an option I'm affraid as it is discontinued but try a free XLProfiler or cheap Profile Mechanic - Scanner from Digital Light and Color (www.dl-c.com).

[mendel_leisk]

Hi Pawel, thanks for the tips.

 

I downloaded and tried XLProfiler. It requires a bit of effort, being freeware. I did manage to get it interacting with Excel and Photoshop, per the intstructions. After a bit of struggle.

 

I did make a profile. The result was extremely "poopy", but that could be due to a number of factors. I made no adjustments, was using a 16bit linear scan of the target, etc, etc. The object was just to see if it would work. I will try experimenting some more.

 

I've heard also that Vuescan's profile making process is pretty basic.

 

Anyway will post if I make any progress with this XLProfiler.

[pawel_bartuzi]

My profiling experience is limited and I admit I didn't try XLProfiler, I only heard good things on DTP newsgroups about it. If you, however, would like to read on the subject of commercial products please read an article "Measuring the Quality of ICC Profiles and Color Management Software" which you can find on the Seybold Report site (an exact link to the PDF with an article is http://www.seyboldreports.com/TSR/subs/0420/0420.pdf).

[mendel_leisk]

Thanks again, Pawel. I exerimented a bit more with XLProfiler, and found that my first attempt was totally off the mark. I had not actually calculated anything based on my target scan and accompanying info file. I had loaded them, but had not started the process.

 

Like a lot of freeware, I've found XLProfiler to have a bit of a learning curve, and it was a bit unruly (it uses MS Excel, and locked up once).

 

Eventually, after trying different things, I did manage to make a proper profile.

 

As I'd used my 16bit linear output from MSU as my source, I tried applying it to same type of file, within Photoshop. Result was quite close. STILL, I liked the results I got doing Vuescan scan-from disk, with it's (admitedly simple) scanner profile. More experimenting in order?

 

I think I will try other software also, perhaps the other one you mentioned. And will check out the reading suggestion.

[mendel_leisk]

Back to my Vuescan workflow, I believe I've improve the color balance further by using manual setting, instead of neutral. The neutral red/green/blue values arrived at by various right-clicks in preview, doing scan-from-disk from the 16bit linear Minolta Scan Utility scan of my Provia target, on it's gamma patch #9, which (along with the entire surround of the target) appears to be neutral grey.

 

I did the right-click procedure at perhaps a half dozen locations, fed the various results into an Excel spreadsheet, averaged them, and entered the averages back into the Vuescan neutral red/green/blue fields.

[pawel_bartuzi]

The procedure you found out to calculate "neutral" white balance within Vuescan sounds interesting (I didn't thought I could use Excel to average white point values). However, manually correcting my scans was exactly the thing I wanted to avoid. :-)

 

The key question is: why profile the scanner? The reasonig behind profiling my scanner was I wanted my scans to resemble the originals as much as possible - hence the calibrated monitor and a lightbox as a tools to do the evaluation of results. Now when I say that now I am content with the results I get using Colortune profiles it doesn't mean the images I get are perfect but rather that the colours on the scans are identical to those on film - even if it means that all exposure and colour cast errors from the slides are visible on the scans.

 

It was next to impossible to achieve that using profiles generated with Vuescan, so I moved to other profiling tools. If you, however, don't mind occasionally adjusting white balance then Vuescan profiled output can serve you very well as a starting point from which only minimal manual corrections should be neccessary to get the image you want.

[mendel_leisk]

Pawel wrote, in part:

 

"Vuescan profiled output can serve you very well as a starting point from which only minimal manual corrections should be neccessary to get the image you want"

 

That precisely sums up my intent!

 

I felt that the Vuescan scanner profiling, coupled with it's "neutral" color balance setting, was quite close. Still, white highlights were a "little" pink.

 

Don't know how familiar you are with Vuescan's right-click on the preview pane. It switches color balance (from whatever) to "manual", and set's neutral (mid-tone?) red, green and blue values to make the spot you clicked 128/128/128. I THINK...

 

Now all I did was run through this procedure multiple times, on supposedly neutral gray patch of my scanned target, WITH the previously produced ICC profile in use. I figure this should "nudge" me closer to neutral, if there were shortcomings in the icc profile.

 

Nothing magical with MS Excel. I just entered all the values, from the various right-clicks, averaged them, then typed the result back into the 3 vuescan manual setting fields, for red/green/blue. It could just as well be done on paper or with calculator. It's just the fact that the target is an analog film, and while the extent of that patch is close to neutral grey, the results vary every time you right-click, becuase you're never hitting the same pixel twice.

 

I suppose the only way to ABSOLUTELY MATCH the colors of an old slide scan to THE ORIGINAL SCENE, would involve a grey card (at the least, and maybe an IT8 target also) and a time machine ;)

[pawel_bartuzi]

<i>Don't know how familiar you are with Vuescan's right-click on the preview pane. It switches color balance (from whatever) to "manual", and set's neutral (mid-tone?) red, green and blue values to make the spot you clicked 128/128/128. I THINK...</i><p>

<p>

I know and use Vuescan right-click procedure every time I scan from disk negative film, I didn't however thought I could enhance it with averaging numerical values of every click.<p>

<p>

<i>I suppose the only way to ABSOLUTELY MATCH the colors of an old slide scan to THE ORIGINAL SCENE, would involve a grey card (at the least, and maybe an IT8 target also) and a time machine ;)</i><p>

<p>

Time machine would be handy... But speaking about matching the scan to the original scene we enter a subject of <i>film</i>, not <i>scanner</i> profiling. :-)