Scanner densitometry; TMX

Discussion in 'Black and White' started by michael_young|3, Mar 23, 2004.

  1. I sandwiched an IT8 to a cut down sheet of 100TMX, and shot the attached. The intention was to plot some development curves using the scanner as a densitometer. Visually, though, the shot has some problems. The shadows are very cleanly separated, but the shoulder is very broad and flat. Is it simply overexposed? A larger, unmanipulated negative image is at: Would appreciate your comments.
  2. Using a scanner as a densitometer is theoretically possible but two related problems must be overcome. I suspect these problems are at work in the example you have submitted.

    1) The scanner itself must be calibrated against known density values before using it to measure unknown denisities. You can scan a Kodak or Stouffer step wedge, but...

    2) In order to optimize scan quality, software for many consumer-type scanners cannot disable automatic exposure adjustments. These will undo any calibration.

    These unknows must be controlled or you cannot reliably use the results from ordinary scanners to evaluate your film. I have experimented with using a scanner as you propose but have not been to produce results that match those from my densitometer. I've gotten close enough that it seems with reach but have not spent the additional time required to solve the problem. Perhaps others will suggest a solution.
  3. The scan is as accurate a reproduction of the negative as my equipment will allow. Auto-exposure and other distortions are turned off. The thumbnail above was converted to grayscale in Photoshop, and inverted to produce a "normal" view. The larger scan, link above, is the unadulterated scan.

    I exposed the frame five stops above what the light source metered, putting it on Zone X. That seemed to be the right thing to do. Does the rest of the frame look OK? The film-base, visible at the left, isn't blown out or too dark. The shadow densities, though, look a bit suspect.

    I would suspect the meter calibration, but I use it expose color trannies regularly with good results.
  4. Maybe it wasn't too far off the mark. Here's a rough plot.
  5. Oops. Meant to resize that smaller... Here's a 400 px wide plot.
  6. Try scanning the IT8 film directly as a reference using the same settings and procedures you used for the TMX film. Since you know the density values in the monochrome steps, if your measurement and calulation methods are correct they should yield the values of the steps. If you get different values (and I suspect you will), you may possibly use them as a rough calibration reference to convert the TMX scan results to actual density values.

    You may also need to repeat the test with the IT8 film at different times to be sure that variation in lamp brightness and other factors are not in the mix. (Fluorescent scanning lamps vary in brightness with temperature among other things. With auto-calibration turned off, the compensation is likely to be lost.)

    How did you calculate the film density values in your graph?
  7. Mike, your curve doesn't look right. I think you are fighting an uphill battle here. The reasons densitometers exist is to obtain reliable results and they are only so accurate at that. With a scanner, I don't think you can ever be sure.
  8. Both the original IT8 and negative scan are in a folder. Development and scanning details are in the comments.
    The chart values came from comparing and interpolating the test negative against the presumed 1/2-stop steps in the scanned IT8 gray step wedge. The IT8 calibration data contains unambiguous CIE- color values for each patch. This would be a more direct translation in Photoshop, and might try this next. (I'm giving up on manual measuring and processing. See below.)
    Hi, Stephen. Aye! and verily! What can be simpler than measuring 21 spots on a negative, and plotting the values? I have a scanner and software development tools, but no densitometer. And so, I asked myself: How hard can that be? Maybe the next guy to come along will mumble to hisself: "I have a scanner and this cool software. Who needs a densitometer?" It needs doing; I'm going to do it.
  9. Mike, how about having the negative read by a lab with a densitometer and then use that data to confirm your scanner's response?
  10. Stephen: Yup. The local community college has one in their darkroom. I enrolled in a course so they'll let me use it.
  11. I suspect that the grayscale steps on the IT8 are not uniform. I don't have one here to measure and could find no definitve answers on the web. In any case, the direct IT8 scan that you posted shows that the steps are not uniform. Whether that is due to the IT8 itself or your scanner and software remains to be determined.

    While it will be possible to use an IT8 reference for B&W film testing, you will need to know the actual densities of each step. The remaining issues of scanner and software effects can then be addressed.

    I put a 4x5 Stouffer step wedge on an Epson 1650 last night (transparency mode) and used the density-reading feature of a program called VueScan. The readings were surprisingly close to the actual values. I'll look into it further this weekend.
  12. Here is the grayscale portion of the reference data. Photoshop displays CIE-L*a*b values as whole integers. The luminance values match within the expected integer round-off. The color component values are very low saturation, and are pretty uninteresting rounded off, but spot checks are also within the display round-off.
    I half expected the high density values to roll-off due to scanner clipping. If they did, the difference would be lost in the integer round-off.
    I measured this on the 16-bit TIF from the original scan. I expect the posted JPG will have some noise and slight shift, but it won't be too far off.
    Spot checking the color patches, not included in the snippet below shows some variance, but overall is within 1 or 2%, usually much less. I'm confident the scanner is reading right, and that the scan settings are right.
    That said, I tried a few different ways to import a B&W negative scan. The worst turned out to be letting SilverFast invert the values. The best is tweaking the negative's luminance values and gamma levels in Photoshop before inverting. A positive of the IT8 inverted this way looks as it "should". I haven't had a chance to rescan a real image yet, but will post a few later if they show a worthwhile improvement.
    PS: I also included a randomly selected portion from the color patch data, in case you were curious and wanted to probe around a little. I'll also compare these values against the TMX scans when I have more time.
    ORIGINATOR "Monaco Systems, Inc."
    DESCRIPTOR "IT8.7/1 Reference Data File, 4x5 Fujichrome"
    CREATED "December 12, 2002"
    MANUFACTURER Monaco Systems, Inc.
    PROD_DATE "2002:12"
    SERIAL "Batch Average Data"
    MATERIAL "Fujichrome"
    #ID X Y Z L A B
    E12 29.8 38.46 4.27 68.36 -25.56 70.91
    E13 44.72 55.18 56.24 79.14 -23.06 -11.98
    E14 45.35 36.61 42.43 66.98 31.16 -17.17
    E15 60.73 64.8 24.1 84.38 -4.07 40.36
    E16 22.56 23.71 19.72 55.8 -1.38 -0.34
    E17 39.33 31.66 17.78 63.06 30.02 16.4
    E18 39.31 50.41 26.51 76.32 -27.17 22.17
    E19 28.5 28.78 44.23 60.59 2.95 -30.43
    E20 1.99 2.98 3.9 19.95 -17.82 -10.32
    E21 2.09 1.46 1.97 12.34 17.28 -8.7
    E22 4.28 4.49 0.91 25.24 -0.74 26.62
    F1 3.06 3.83 2.56 23.1 -10.24 4.55
    F2 2.65 4.16 2.21 24.18 -22.29 9.44
    F3 2.32 4.35 1.96 24.78 -31.49 12.83
    F4 2.28 4.73 1.92 25.96 -37.3 15.25
    F5 10.66 13.8 8.81 43.94 -18.41 8.45
    F6 8.66 13.93 7.52 44.13 -35.25 13.64
    F7 7.49 14.21 6.36 44.54 -47.63 19.24
    F8 7.73 14.75 6.06 45.29 -48.58 21.89
    F9 27.78 32.23 22.95 63.53 -12.57 6.56
    F10 24.55 32.29 21.44 63.58 -26.12 9.57
    F11 21.58 31.77 19.1 63.15 -37.59 13.66
    F12 17.47 28.92 13.78 60.71 -47.74 22.12
    L22 0.44 0.47 0.91 4.28 -0.85 -9.55
    Dmin 66.89 70.89 56.92 87.43 -3.2 1.6
    GS1 51.65 54.16 43.95 78.55 -1.48 0.89
    GS2 45.68 47.95 38.19 74.8 -1.57 1.83
    GS3 40.19 42.34 33.86 71.1 -1.94 1.54
    GS4 35.14 36.82 29.22 67.15 -1.25 1.84
    GS5 30.17 31.7 25.13 63.09 -1.47 1.8
    GS6 25.81 27.11 21.24 59.07 -1.36 2.2
    GS7 22.01 23.19 18.21 55.27 -1.63 2
    GS8 19.22 20.27 15.98 52.15 -1.65 1.78
    GS9 16.08 17.04 13.37 48.31 -1.96 1.84
    GS10 13.27 14.04 11.23 44.29 -1.76 1.06
    GS11 10.83 11.47 9.11 40.36 -1.68 1.21
    GS12 8.81 9.4 7.35 36.74 -2.1 1.62
    GS13 6.91 7.36 5.87 32.62 -1.91 0.96
    GS14 5.23 5.59 4.39 28.35 -1.89 1.25
    GS15 3.93 4.2 3.39 24.33 -1.81 0.5
    GS16 2.97 3.16 2.55 20.67 -1.35 0.47
    GS17 2.14 2.29 1.84 16.93 -1.41 0.52
    GS18 1.43 1.53 1.22 12.81 -1.44 0.57
    GS19 0.9 0.97 0.75 8.75 -1.3 0.99
    GS20 0.42 0.47 0.36 4.22 -1.21 0.52
    GS21 0.23 0.25 0.2 2.28 -0.67 0.14
    GS22 0.13 0.14 0.12 1.26 -0.2 -0.01
    Dmax 0.09 0.1 0.09 0.89 -0.07 -0.08
  13. I too saw the data chart posted above when poking around for some info on the IT8 film. Unfortunately, as I currently understand it, the values in the chart alone do not yield density of the IT8 grayscale steps. The L value in an L-a-b triplet will also depend on gamma used in the monitor/software/scanner as will the log transmission density.

    You've probably already checked out the following sites but in case you haven't:

    These sites illuminate the relationship between L, gamma, and log transmission density in a computer/scanner measurement system.

    I agree that it would be great to use a common scanner as a densitometer for evaluating B&W film. It's a worthwhile project and would make a useful tool.
  14. Thanks for the links. The gamma would be 2.20, that of AdobeRGB, the working space when I imported the scan. It might also be sRGB, also gamma 2.2, very likely the color space Photoshop's Save for Web utility wrote. In either case, 2.2 would be a good starting point.

    It's worth looking into. I hadn't thought the referece data could be translated to density values directly. I'll try that in the next day or two.
  15. My best results from using a scanner as a densitometer were to assume zero density to be pure white on the screen, which reads 255 in Photoshop LE Level Adjust. Read it with the probe without Preview. Other densities are log (255 / probe reading). The high densities are the least precise because I can only read integers between 0 and 255.

    It's not bad for estimating the density scale of a negative, but not very good for comparing H&D curves. IMHO goes with all my statements.

Share This Page