There have been a few threads here in the past asking how to use a scanner as a densitometer. Someone suggested me a source of an inexpensive set of films of known densities. I have lost that information and can't resume the project.

My scanner can scan transparent material of fairly large sizes so I can use Kodak density tablet, but if I try to exchange data with those who use 35mm scanners I need to find a micro tablet that fits in their scanner... (Besides locking exposure and getting the raw data from the scanner saved in a TIFF file, which my system can easily do.)

The idea is to fit a smooth function (perhaps polynomials or continued fractions) into response pattern from scanner using step tablet to do interpolated modeling, and then invert that model to estimate density from other materials. I prefer to have multiple sets of tablets of known densities so that I can verify the model accuracy without reusing the data that created the model.

For those who use 35mm scanners might prefer to include many steps in one frame so that they don't have to feed many frames to get one curve. This is only an experimental thought. How do people here do densitometry? One graycard and modifying exposure in each frame using shutter/aperture/ND filter/etc? I think contact printing step tablet is a good way but again I don't know a step tablet in 35mm size... maybe photographing reflective density tablet?

I am looking for product information (including information for free samples for research purposes :-) that can potentially be useful for these purposes.

-- Ryuji Suzuki (rsuzuki@rs.cncdsl.com), May 20, 2002

Answers

Look at http://www.stouffer.net/Photo.htm

-- John Hicks (jhicks31@bellsouth.net), May 20, 2002.

Hi Ryuji. I believe you might be referring to my recommendation of Lee Filters as a source of free samples of ND gels. They no longer advertise the sample swatch books on their website, but I'm sure if you contact them, they'll do their best to oblige.

If a scanner puts out truly RAW data, then the response should be linear. Any noticeable non-linearity will be down to offsets in the A/D conversion chain, together with light flare and possibly IR 'fogging', or breakthrough, in the higher density region.
If the 'raw' data are nowhere near linear, then the scanner software will have translated the native linear response of the CCD to a gamma function (most likely 2.2 or thereabouts), again with the inclusion of any offset. The only difficulty in measuring this gamma curve is in finding the zero density point, or white point, of the scanner. It's usually set somewhere around 0.1 ~ 0.15D absolute.
Once that's established, the transfer function of the scanner should be reasonably easy to graph.
Nearly every scanner autocalibrates its whitepoint with every scan, so repeatability should be checked over a period of time, and over a reasonable range of temperature.

The problem of infrared punching through the dyestuff of any sample under test might not be trivial. The problem can be detected by sampling a known silver film image density, or a reference thin metal film density wedge, and comparing those readings with known densities of dye-image sample.

-- Pete Andrews (p.l.andrews@bham.ac.uk), May 21, 2002.

I have found by using semi-log graph paper that the scanner output numbers as obtained from Photoshop level adjust pointer plot as a straight line against decsity of a scanned step density tablet, either transmitted or reflected. Plot the scanner reading on the log azis.

-- Patrick A. Gainer (pgainer@rtol.net), May 21, 2002.

John - thanks for that link. It sounds like they come in enough variety to choose from.

Pete, yes, I meant your previous info, which I couldn't find... even through google with site: specified.

The raw data I get is set for gamma of 1.0 with some unknown brightness and contrast. (note: gamma here is a different stuff from those used for contrast for those not familiar) So it should be pretty linear to light intensity (whether transmitted or reflected) except for the effect of noise, device nonlinearity, etc. However, these are easy to fix as long as I have a good calibration material.

For those who are confused by "log" and "linear" thing. The linear here is meant for intensity-value relation. Strictly speaking, it's not linear because quantization is a non-linear function... I'm sure there are some DSP-heads around here. Gamma in the context of CRT, printing device, etc refer to the exponent g in intensity = (contrast * value)^g + offset (or something more accurate than this). Optical density is defined by common log of ratio of incident and transmitted light intensities. For some reason, I like playing with slide rules but I haven't used log scale plotting paper for years.

-- Ryuji Suzuki (rsuzuki@rs.cncdsl.com), May 24, 2002.

"note: gamma here is a different stuff from those used for contrast for those not familiar" - Actually, it's substantially the same thing Ryuji.
Since sensitometry curves are all plotted log/log, even the 'straight line' portion of a film curve follows a power law (gamma function). Only when film is developed to a gamma of one is it truly linear.
A gamma of 0.7 on a film sensitometry curve has exactly the same meaning as a gamma of 2.2 applied to a computer monitor. It's just that when converted back to a linear form, a gamma < 1 means that the curve has a positive hump, while a gamma >1 has a negative dip - that's all.

As an interesting aside to this: If we developed our negatives to a gamma of 0.45, it would almost exactly cancel the (sRGB) gamma of 2.2 of a computer monitor, and we'd get a linear translation of brightness from subject to output device.
(This is probably why I prefer to have my monitor set at gamma 1.8; the reciprocal power of 0.55 is much closer to a normally developed film gamma.)

-- Pete Andrews (p.l.andrews@bham.ac.uk), May 24, 2002.

Thanks Pete, that was my bedtime posting (and probably morning for you) but I meant to say more subtle differences, and confusing term "contrast" used with CRT, which is more like "speed" in film. I guess we should start another thread on this topic maybe when I get my densitoscanner working...

I prefer to scan at higher precision and numerically manipulate the gamma because changing development also changes things like accutance and granularity. (numerically changing the gamma also affects but at least print is the same... which counts more to me)

-- Ryuji Suzuki (rsuzuki@rs.cncdsl.com), May 24, 2002.

Ryuji, You should revisit semi-log and log-log graph paper. It really is a big help when you are seeking a curve to fit data. If you scan a step density tablet and the numbers obtained from PhotoShop level adjust with preview shut off plot on semi-log paper as a straight line, you have a pretty good idea about what kind of erquation will fit. If you want to skip the equation, all you need do is scan your negative, use the same level adjust function to read salient points from the neg, and then read density values from the graph. You can apply the same procedure to read reflection density from prints by scanning a reflective gray scale and plotting the level values of the steps. Of course, you do not really need to use semi-log paper, but it does help to stretch out the increments between low density steps.

-- Patrick A. Gainer (pgainer@rtol.net), May 25, 2002.

Patrick, I just hate manually picking up points from the film image and manually plot things. I'd rather read values automatically, and plot curve out of a laser printer with relevant statistical accuracy measures indicated. I also want to store numerical data in data file for later analysis.

Mathematicians don't deal with dirty numbers. Applied mathematicians and electrical engineers use numerical computation and plot on screen or print out. I even saw an EE graduate student who used semi-log paper upside down without knowing it -- it's just not used as widely as before. Most engineering students don't know how to use slide rule... maybe they can do logarithm and multiplication if you quickly explain, but that's not a part of "using" it. It is just that I do not do things in your way... I don't even have a Photoshop.

-- Ryuji Suzuki (rsuzuki@rs.cncdsl.com), May 25, 2002.

Ryuji, Does this mean you are not wanting to know, in an artistic sense, where in a photographic image you want black and white and various in between shades? The computer will do it for you? Yes, you could write software that will do what most photo editing software does--sort through the image and pick out maximum and minimum. I am not usually happy with those results. I also do not own the full PhotoShop, I have the PhotoShop LE that came with one of my scanners. I still think you would learn to have fun with log paper. What young engineers do not know about old ways is a shame. I went to our local college library one day because I fell heir to an old oscilloscope that needed repair. It contained vacuum tubes that were not in any of my antique manuals, and I foolishly thought the college might have a manual somewhere in the archives. When I asked at the desk where I would find such a book, the attendant said "What's a vacuum tube?" Oh, well.

-- Patrick A. Gainer (pgainer@rtol.net), May 25, 2002.

Patrick, that's not too shocking. Many young EE people can't even read color codes on resistors. When they pick up a capacitor saying 473, they don't know what it means. There are many these stories. To those people, you don't show them something saying "12AX7" and expect them to know what it is.

Let's talk about photo-related thing. The population that even recognize silver halide cameras is shrinking rapidly. Even those who know EOS-1 or F5 or something, many don't recognize Copal metal square shutter or #1 lens shutter as a shutter mechanism! You show them how mechanism works, and explain the self timer unit compared to old mechanical wakeup alarm clock. Don't expect they know what you mean!

There are many many many of these.

We just don't need to know them.

I have a mini collection of old tube devices and a boxful of tubes. My old university wrote off many of those measurement devices in the lab and I took some of the nice ones. These days, people are getting rid of CRT's, leaving only magnetrons in the lab's kitchen. However, I use automated data acquisition, recording, etc. for my research. For the same reason, I wrote a postscript program to make semi-log paper out of the laser printer, but I don't use it for actual work. I either plot data electronically, or I just pick up a sheet of scrap paper, draw log graduation with a pencil - 1.5 digit accuracy is often good enough.

Please don't think I just don't use Photoshop. I should have said I have the same LE version of Photoshop, which came with the scanner, but I never installed it. I don't use a Microsoft product nor an Apple product.

Please don't tell me to convert to Microsoft platform unless you are prepared to be told opposite.

-- Ryuji Suzuki (rsuzuki@rs.cncdsl.com), May 25, 2002.

Ryuji, It's too bad we cannot be in the same room arguing with each other. I think we would enjoy it.

-- Patrick A. Gainer (pgainer@rtol.net), May 26, 2002.