Scanner as Densitometer - Linearity Issue

[hchapman]

A note to those using flatbed scanner measurements to derive correction curves for your negatives. Ive obtained non-linear results when using an Epson V700 scanner to measure a calibrated Stouffer 21 step reflectance step wedge (Stouffer #R2110C). When the step wedge was measured with an Eye One spectrophotometer the measured density readings matched the reference densities very well. When the step wedge was measured with the scanner the results were markedly different. The scan was performed as gray scale with all input/output controls set to 0/255. The scanned file was read in photoshop after using levels to set the black step to max black and white step to max white. Nearly identical results were obtained from this scanner using the stock scanner profile and a new scanner profile prepared using the Eye One. I also checked a friend's V700 using the Stouffer step wedge and got similar results. Both scanners function normally for general scanning purposes.
When the scanner is used to prepare a correction curve that, applied to a step wedge and printed, reads as linear with the scanner, the same print read with the Eye One shows a curve that looks like the inverse of the curve from reading the Stouffer step wedge with the scanner. If one's calibration standard for linearity is the Stouffer wedge, prints prepared using V700 scanner-derived curves may not be linear. For what it is worth, which may not be much.
I've seen instructions for using scanners to prepare profiles and curves for print linearization on this forum, in the "Digital Black and White - The Print" forum, and from other sources. I'm curious if my results have been observed by others and if they are relevant to preparing negatives.
Graphs of readings from the Stouffer step wedge with the scanner and Eye One are attached.
-Harlan

 

[Ron-san]

A note to those using flatbed scanner measurements to derive correction curves for your negatives.
-Harlan

Harlan
The steps on the Stouffer grayscale are in even increments of reflected optical density. According to my dim understanding of the subject, there is no particular reason why those density steps, when plotted against K value (%gray) should yield a straight line.
To see if your scanner is reading reflected values accurately, take a look at Paul Roarks website ( Dead Link Removed ). If you drill down into his site you will find an article on using a flatbed scanner to make ICC profiles for B+W. In the intro to this article is a good section on calibrating your flatbed scanner. After reading your post, I decided to test my two scanners (an older Epson 1680 and a newer V700). As Paul suggests, I dug into my drawers until I found an IT8 grayscale reflectance target that came with a Monaco EZ color program that was bundled with the 1680. In his article Roark gives the Lab values that various patches of that grayscale target should read. For example, Step 0 should read 92.5, Step 9 should read 51.7, Step 18=15.7, Step 20=7.9, Step 22=5.9.
I put the grayscale on the 1680, scanned it, and opened the scan in Photoshop. I set one of the readouts of the Info palette to read Lab color, which meant that when I put the eyedropper tool over a scanned patch, I would get a readout of the Lab value of that patch.
Then I opened the Levels window. I moved the black point slider and the white point slider around until I got a Lab value for Step 22 of 5 (close as I could get to the target value of 5.9) and a Lab value for Step 0 of 93 (close as I could get to 92.5). With these end points set, I used the eyedropper to measure the intermediate points. Step 9 read 52, Step 18 read 16, and Step 20 read 7. Those readings are very close, in my estimation, to the target values of 51.7, 15.7, and 7.9.
So, somewhat to my surprise, I conclude that my venerable old 1680, once I have the black and white endpoints set, is reading Lab values with considerable accuracy.
I repeated the experiment on the V700. This time, after scanning, opening Levels, and setting the Step 0 and Step 22 endpoints, Step 9 read 51, Step 18 read 14, and Step 20 read 6. Steps 18 and 20 were thus a little bit out of line. To fix that, I grabbed the middle slider of the Levels window and slid it to the left just a teeny bit. I quickly found a setting that caused Step 18 to read Lab=15 and Step 20 to read Lab=7. Right back on target.

Conclusions: I had always assumed that my flatbed scanners were not particularly accurate, but never checked it because the profiles and digital negs I made using them worked very well and made good prints. Now that I have gone back and checked them, I find that they are accurate beyond all previous hopes.
If that slight inaccuracy of the V700 bothered me I could rather easily fiddle with the adjustment curve tool in the Lasersoft driver until I came up with an adjustment curve that would cause the scan to give dead on Lab readings. I expect I will not bother because my printing procedure has much greater errors elsewhere in the system.

I know there are folks out there who understand this material to a much greater depth than I do. I would be interested to hear if what I have written here makes sense or is just the usual delusional.
Cheers, Ron-san

 

[hchapman]

Scanner calibration progress

Hello Ron-San,
Thank you for your thoughtful reply.
I've made progress working through Paul Roark's instructions. Have found scanner settings that will produce a scan that, read in Photoshop, gives Lab values that are workably close to what the Eye One reads. I'm impressed by the result. The next step will be to see how the grayscale K values look, will let you know how that falls out.
-Harlan

 

[holmburgers]

I'd like to revive this topic, and pose this question:

Is it necessary to buy a densitometer if you have a scanner?

 

[donbga]

I'd like to revive this topic, and pose this question:

Is it necessary to buy a densitometer if you have a scanner?

Just read Harlan's post for the answer.

 

[Ron-san]

My belief is this: Not only is a flat bed scanner adequate for calibrating digital negatives, I actually do no know how to use a densitometer for that purpose. In the process of making a so-called "gray curve" (a Photoshop .acv curve) to linearize print tones using QTR, you go through several steps. First, you read the K values of various steps from a Step Tablet print, then you transform those values into negative space, then you invert them to make a correction curve. I know how to do all that with the %gray (K) values I read using a scanner and the Photoshop eyedropper tool. But if I were to read the actual reflective densities of the steps, using a densitometer, I do no know how to transform those readings into negative space nor do I know how to invert them to make a correction curve. Until someone with more mathematical skill than I shows me how to do that, I will of necessity keep on with the flat bed scanner.

On the positive side (no pun) I have been using an Epson V700 flat bed scanner for years to make QTR profiles for making negatives for various alt processes. The resulting prints look fine to my eye and have passed the test with various galleries, museums, and assorted photo experts that I have shown them to. So I feel, by actual real world test, that the flat bed scanner method works quite well.

Recently I had to work with a new scanner, which unknown to me, had an auto function that moved the midpoint of the histogram to some value decreed by the machine. QTR profiles produced with that scanner made awful prints, even though I was doing everything else correctly. And when I found the auto bug and turned it off, the profiles and prints came out just fine as usual. So I am gaining confidence that even these low end scanners do a fine job for our alt printing purposes.

Cheers, Ron Reeder

 

[holmburgers]

Don, Harlan's post seems very telling, but the truth is.. I don't know what the Eye One spectrophotometer is. Nor do I understand exactly what he means with "Nearly identical results were obtained from this scanner using the stock scanner profile and a new scanner profile prepared using the Eye One."

I think a lot of my trouble with these discussions is that everyone is using Photoshop, I'm using PaintShopPro9. I realize that I'm driving the Yugo of the photo editing suites, but it's surprisingly adequate for my purposes.

Ron, given my previous statement, I'm lost when words like "negative space" come up. Is that from A Brief History of Time?

Are you guys talking about calibrating digital negatives by scanning them and implementing a correction curve?

Truth be told, I'm curious about how to use a scanner for densitometry readings of analog negatives.

 

[Ron-san]

OK, seems you are starting from farther back in the bus than I thought you were. The original thread concerned whether inexpensive flat bed scanners are linear enough to calibrate digital negatives. And I think the answer is yes, partly because the calibration process involves using the scanner to read reflection densities over a relatively modest range (densities of about 0-1.5 for pt/pd, maybe 0-2.1 or so for silver gelatin, etc, etc). The Epson V700 can handle that dynamic range.

But apparently your question is whether such scanners are useful for reading the transmission densities of analog negatives. Analog negs often have a transmission density range considerably greater than the V700 can handle (despite the mendaciously optimistic claims of the manufacturers). So, forgetting about questions of linearity, the V700 is not much use for analog negs unless they are deliberately under-developed to decrease their contrast range.

Is that answer on target? Cheers, Ron Reeder

 

[Bob Carnie]

Chris

I have an eyeone, a densitometer , and a couple of scanners.
for reading step wedges for colour control and plotting colour charts we would use a densitometer to do this and apply specific numbers to reference number.
For producing profiles for papers we use the eyeone spectrometer, and it is also to my knowledge able to give L readings which I will speak about.
Ron has clearly pointed out how he uses a scanner to produce curves for his printing processes { he was in Toronto and graciously showed my staff how to do this}
Then there is the info Pallette in PS
Some use RGB numbers, Some use K numbers and Some{myself} live and die with the L numbers.

I grew up colour correcting on enlargers and the L Mode and number system is the most intuitive system for me, I love the way the A & B numbers work and since the L channel is a 0-100 system it resonates with me after reading and tryint to understand writings by Ansel Adams, Fred Picker, Phil Davis on the zone systems for black and white expansion and contraction

Since the eye one spectometor works in *L mode* I think I will graduate to this device and learn all I can. This way when using PS and reading numbers on paper or film I can understand the numbers more clearly than reading RGB numbers.

Someone like PE over on APUG would probably be more comfortable with a Macbeth Densitometer as probably many here.
Some of the Platinum workers I have met,, Kerik, Ike and Sandy use the scannner much like Ron Describes but also all are very smart gentlemen and understand how the densitometer works and plotting film and paper curves, I am not that smart yet.

So all these devices are tools that you can use, and depending upon your background and likes/dislikes the right tools will become obvious.

I think the eyeone spectometer is for you but to start I would go for the Macbeth Densitometer{$ 300 uses}.

You are very interested/investigative on many process fronts so I believe a way of reading the numbers is in the cards for you.. You can borrow time on labs equipment in your area if you are polite enough, to get started and see what way most suits you.

Bob

 

[holmburgers]

Hey guys, thanks for the responses.

It's true, I'm sitting way in the back of the bus when it comes to the D in DPUG. I have an Epson 4990 and that's the extent of my "hybrid" workflow at this point.

Ron, what you say about density makes a lot of sense. Indeed, for that reason alone I'm sure a scanner cannot replace a densitometer. Afterall, I get noisy scans with a slightly underexposed slide.

Bob, yes, the densitometer route is probably best for me now, particularly to learn the "art" of the Hurter & Driffield curve. At the moment, I'm trying to make sense of a dye-imbibition process that I'm working on, and needing to understand my input & output values (all analog). The ability to understand the old school sensitometry language of characteristic curves & the like and then relate this to the # values from the tools you speak of sounds like the best way to get a "comprehensive eduction" in the topic.

The thought of digital negatives is becoming very appealing, particularly for this process. Ultimately, I look forward to doing this someday. The problem at the moment is that to get the same quality digitally that I can get optically/analog is too expensive for me.

Again, I appreciate the advice.

CHeers!

 

[Bob Carnie]

A very long, lifetime learning curve btw.

 

[holmburgers]

I'm cool with that...

 

[indigo]

I wonder how to use a film scanner as densitometer. I want to measure film density. What the pixel values related to density? I have the Konica Minolta Dimage Dual Scan IV. I also have a color transmision/reflection densitometer. I know how to use the desitometer well but don't know how to use a scanner as densitometer.

 

[clay]

I'm going to go all propeller-head on you.

The commonly used %black that photoshop uses comes from the printing world. It is also called dot percent. Basically it is a measure of the amount of coverage of black ink over each little unit (think pixel).

Reflection density as measured by a densitometer is an indication of the amount of incident light falling on a subject that is reflected back. It is usually given in log base 2 terms for mathematical convenience. So the 'blacker' something is, the less light it reflects back. So a 1.8 reflection density means that 1.8 / .3 or 6 stops of light loss occur between the incident light amount and the reflected light amount.

The relationship between these two ways of measuring is a formula called the yule-neilsen formula. Google it and read about if you are truly curious. You will notice that a constant is used as a plug. This immediately raises some suspicions when you see it in a formula in my experience. In this case, it is used as a plug to compensate for the apparent perceptual differences between glossy substrates and matte substrates.

In other words, you still need to use your eyes. I have a spreadsheet somewhere that I will post here that will convert logD reflection density to dot percent. Use it as a guide, but realize that changing that constant can have an effect on your results.

It is easy to get all tangled up in the numbers and lose sight that ultimately it has to look good. Make that the goal, not numerical perfection.

 

[holmburgers]

clay, that's a very informative post.

Do you mean log base 10 though, or am I missing something? (this wouldn't be entirely surprising)

I guess that if you calibrated your whole process using a scanner only, perhaps you'd be just fine. But trying to relate scanner readings (%black/dot percent) to densitometry readings might get you into trouble. That's what I'm taking away from this.

As for reading transparency densities, that goes back to the "transmissive" capabilities of scanners, which might not be up to the challenge.

 

[clay]

Yes, sorry, log base 10.

My main point is that most people are better off sticking to the scanner approach in conjunction with their own eyeballs rather than trying to finesse all the trickiness in converting from reflection density to dot percent. That said, for those who want to try:

 

[holmburgers]

I see... for me the goal would be to obviate the need for a densitometer in understanding the characteristic curves of different film, developer and print techniques. That being said, I will be looking for a densitometer now...

Thanks for posting the worksheet

 

[Bob Carnie]

Try to find one that does transmission and reflection readings. good luck

I see... for me the goal would be to obviate the need for a densitometer in understanding the characteristic curves of different film, developer and print techniques. That being said, I will be looking for a densitometer now...

Thanks for posting the worksheet

 

[clay]

The Heiland TRD 2 does both. I have had one for about a dozen years with nary a problem. It does not do UV, however. Visual spectrum only.

 

[michaelbsc]

... It's true, I'm sitting way in the back of the bus when it comes to the D in DPUG. I have an Epson 4990 and that's the extent of my "hybrid" workflow at this point....

Don't feel bad Chris. I'm beginning to feel like I missed the bus. I hope they're running on a regular schedule.

MB

 

[holmburgers]

Don't worry michael... I think it's running a non-stop schedule, and the drivers are very accomodating.

That Heiland looks like a beauty, but my fine-wine tastes don't jive with my beer pocket... heck even good beer is expensive!

As is usually the case, I'm looking for an inexpensive, but solid unit. (the Tilt-All of the densitometer world) I'd like one of those old Kodak art deco ones, but I belive they're only transmissive (?). I kind of like the idea of a visual as opposed to photometric densitometer... again, I love old school. I mean, who doesn't want their densitometer to work after a giant EMP?

 

[michaelbsc]

Chris,

In "Beyond the Zone System " there is a description of making one using a spot meter.

Michael

 

[indigo]

I have the X-Rite 810 it's a combination transmision/reflection densitometer and it's a color unit so it's good for both color and black and white. It has been discontinued and was selling for $3500 new but I got it used for $150 in perfect condition so you really don't need to spend a lot for a good densitometer.

 

[holmburgers]

The X-Rite sounds pretty wicked.

Michael, I'd be curious to see that description. I'll have to look for a copy at my local biblioteca