rephotographing film as a means of scanning

Discussion in 'Digital Darkroom' started by chris_autio, Sep 4, 2020.

  1. 80 lp/mm multiplies out to a 2000x3000 pixel image, roughly 6 MP. If you are old enough to have stood near a projection screen, you would know slide film resolution to be ... underwhelming at close range. Youngsters (relatively speaking) will have to rely on a 10x magnifier. This is not to detract from the appearance of Velvia. The colors and contrast are striking, and nearly impossible to emulate with a digital camera. On the other hand, colors in a digital copy of Velvia are true to the original, sometimes better (e.g., shadow detail and white balance).

    The term "line pairs" is misleading. Resolution is measured by examining the space between two parallel lines. Two lines = one space = one pixel.
     
  2. I think that Dave is correct about his Canon 10D resolution. When I look up the pixel size for a Canon 10D, it is 7.38 microns. When I follow the math, and extrapolate the table for the Nyquist limit of resolution associated with pixel size given in,
    Resolution | Edmund Optics
    this pixel size produces about 68 l/mm. For the same sensor size as a 10D, the sensor would need to be (80/68) squared * 6 MB = 8.3 MB to produce 80 l/mm.

    I don't know about Velvia. I only used Kodachrome or black and white in my 35mm film era, which dates me.
     
    Last edited: Sep 21, 2020
  3. The uncertainty (blur) when photographing film is additive by the root-sum-square law. The math gets a little involved when on medium is continuous and the other discrete, but the RSS rule is a reasonable approximation.

    That means a copy of a fine-grained color film like Kodachrome or Velvia with a Canon 10D will have about 1.4x as much blur as either medium taken separately. When one medium has 3 or 4 times the uncertainty of the other, the contribution of the sharper medium can be safely ignored.
     
  4. I’m happy to post a challenge. If Velvia has more than 6mp, I get all of the camera gear of the naysayer. Anyone care to play a game?
     
  5. Just a couple samples. First from 8mp, then from 12mp, then from Velvia 100. What was that about 6mp being a match again!

    8680C25C-5D47-4A4B-A3A5-0645680A0D87.jpeg 90A91970-849E-478A-A77F-A537497BF9DF.jpeg 623E45F0-9EE2-4E96-83A1-AA4C44AD3E42.jpeg
     
  6. Velvia is cited as 160 lp/mm with a high contrast target, as shown above. From this example, 12 MP seems to be a good match. 80 MP is for "normal" edge constrast of 6:1.

    Were the first examples direct digital, or copies of film (Velvia)? I would expect to see individual pixels at this magnificaiton. Do the halos come from sharpening, or sampling of a JPEG image?
     
  7. The Velvia is obviously fuzzier, whether it resolves the lines or not, it has far less acutance.

    I recently tested some T-max 100 with a faded test target, and it could barely clearly resolve 60 lppmm - that's with a lens that easily resolves over 100 lppmm using the same target on a digital camera.

    And where did those samples come from Dave? Are you just re-posting someone else's results again?
     
    Last edited: Sep 24, 2020
  8. A few more questions:
    • What does the, seemingly arbitrary, scale on those curved resolution lines indicate?
    • If it's lppmm, then why is it not marked as such?
    • Why are we not seeing coloured speckles in the Velvia sample? - last time I checked, Velvia was a colour film with cyan, magenta and yellow dye-clouds making up its image.
    • Why have the digital samples been obviously differently processed? - With different sharpness levels for example.
    • How was the film digitised, and at what resolution?
    • Traceability - Were all the samples taken by the same person using the same lens and under identical circumstances?
    • If so; why does the white-level vary between samples?

    With any high stakes bet, it behoves the gambler to ensure that the game isn't rigged!
     
    Last edited: Sep 24, 2020
  9. The discussion was resolution. I can improve acutance by sharpening in post. And who cares where the test comes from? You seem to have this issue in an attempt to divert from the results. I presume when the COVID vaccine comes out, you will do your own tests because citing others makes it false. The film won. It surpassed 12mp. It is better than 6mp. Discussion over.
     
  10. Like any antique, such as this film v digital debate, provenance is everything.

    I'll take your response to mean 'Yes, I've just regurgitated some 2nd hand images that prove nothing."

    That's in contravention of forum rules too.

    Those scale numbers, as I've discovered, mean x100 divisions of image height. So without knowing the format size, they're meaningless as an absolute comparison. And which image 'height'? Short side of frame, or long side?

    The Velvia shot could have come from a medium format shot for all we know. Giving it a pathetic lppmm figure of only 25.
    A typical non-sequiter and ad-hominem argument, made by someone who has no sensible response.
     
    Last edited: Sep 26, 2020
  11. The statement was made that 6mp equals a 35mm Velvia frame. I proved that to be incorrect. Thus, the discussion is indeed over....and welcome to my ignore list.
     
  12. Well maybe not quite over... I'd like to point out that you gave the digital cameras every benefit. If the sensor is in the "normal" orientation its highest resolution is in the vertical or horizontal direction. But diagonally the pixels are farther apart (you expect only about 70% of the resolving power for diagonal lines). Whereas Velvia should stay the same in any direction.

    And the second advantage is that you compared black/white (vs color). Likely your digital cameras uses Bayer filter arrays on the sensor. Such a camera has much lower resolving power on "pure" red and blue colors. So by using a test target with b&w only, you let the digital cameras put their best foot forward.
     
  13. Bill, we don't even know which cameras were used.
    I'm guessing that they were, probably, both older DX/APS models, and with an Anti-aliasing filter over the sensor. This AA filter will knock the possible resolution back by about 20%.

    The devil is in the detail, which is why I was pushing Dave for more of it.

    Mr Luttman, unfortunately, is in the habit of posting other people's example images without acknowledgement, and with little idea of the methodologies used by those 3rd parties. This is extremely bad practise that would immediately disqualify his 'evidence' from any scientific study.

    Gathering images from here, there and anywhere on the internet proves absolutely nothing.

    So, no. Dave has definitely not proven anything - except that he seems unwilling to do his own testing or to engage in a sensible discourse.
     
  14. In lieu of procedural and measurement standards, test results like these are only for comparison purposes. Moreover, we must know what is being compared. If we have three examples made with unspecified methods, and characteristics which are inconsistent with the purported results, comparison is impossible.

    Informal comparisons can be informative, and constitute a significant part of these forums. But we must know that only one parameter is being compared. Everything else must be as consistent as reasonably possible. Save the Latin Square experiments (multi-variable) for the laboratory.
     

Share This Page