Scanning old photographs question

Discussion in 'Beginner Questions' started by tomazdrnovsek, Jun 28, 2020.

  1. I spent many months learning how to 'restore' old photos. Both from courses and from my betters. I don't have much experience in the discussion between 300 vs 600 dpi. I would only say that - for smaller photos - 600 dpi - gives gives restorers the best chance at restoration. Both in terms of clarity (eyes) and in terms of repairing damage.

    So if your photos are big and sharp enough and are pretty much undamaged, 300 dpi is fine. The smaller the photo and the more damage to it, the greater the argument for scanning at 600 dpi.

    tomazdrnovsek likes this.
  2. Tony Parsons

    Tony Parsons Norfolk and Good

    Do you mind me asking, what is the intended purpose and outcome of this exercise ? Is it just to hopefully protect the collection, to share on CD with other family members, to assist in genealogy research - or what ? Just curious.
  3. There are a couple of things to consider aside from image quality of finished product. Can the photos be easily removed from the albums or are they affixed with glue or tape? If they are easy to remove, good. If not, then you may need to lay the entire album page on the scanner bed. Is there information on the back of the photos? If so, don't forget to scan that also. Will your current scanner allow you to place multiple photos on a bed and then scan in batches?

    I've scanned many albums and boxes of old photos. Many others here have done this also, as you can see. A decent flatbed scanner with color correction and multiple file saving capability will probably serve your purpose well. They have a nice array of features that you'll probably find useful. It's a good tool to have and you don't have to break the bank. I found that 300dpi jpegs work well for basic mass scanning of everyday images, such as you mention.. When there is a photo that requires more attention, a decent scanner will allow you to do that.

    You may find that the majority of the work will be prior to scanning. It's helpful to have some organizational scheme in mind and try to label the images in a way that anyone can understand. You're correct in that you have a big job ahead. Best of luck with it.
    mikemorrell and tomazdrnovsek like this.
  4. They are not assumptions. My statements are based on scanning thousands of snapshot prints from past decades. Textured papers such as "silk" were popular after the 1960s because they hid the fingerprints that marred glossy papers. In the 1980s, glossy prints from mass-market photofinishers were the exception, not the rule. Sometimes glossy wasn't even an option or cost extra.

    And though it's true that people have different ideas about camera quality, the original poster specifically said his pictures were made with "not very expensive cameras," which in the 1980s likely were 110- or 126-format Instamatics or even the truly awful Kodak Disc cameras. However, some pictures may have been made with early 35mm autofocus compact cameras like the Canon SureShot series, which were pretty good.

    File sizes are still important. One of my family photo archives won't even fit on a 128GB card or thumb drive. And keep in mind that the flash memory in those devices isn't archival. The electrons will dissipate in a few years and the pictures will be gone. (The persistence depends on many factors; some SSDs are rated for only six months of offline storage.) Large archives like mine are impractical for DVDs, so external hard drives rotated between on-site and off-site locations are the safest solution. Another is cloud storage, but I wouldn't depend on it solely.

    Before starting my archive in the 1990s, I performed many experiments to find the optimal scanning resolutions for various media. Those fundamentals haven't changed. You can scan 4x6-inch Instamatic snapshots at 3200 dpi if you like, but those enormous files won't extract any more image detail than a 300 or 600 dpi scan. This forum is for beginners' questions so I'm trying to offer useful advice for beginners. They shouldn't have to repeat the same experiments to arrive at the same conclusions.
    tomazdrnovsek likes this.
  5. Just to protect the collection and have all my photos in digital form in a cloud.
    Tony Parsons likes this.
  6. I started the process with my current scanner. I decided on 600 dpi JPGs.
    Thank you all for the comments! It made the whole topic a lot clearer to me.
    Of course, I'm still open to more suggestions.
    mikemorrell likes this.
  7. Tony Parsons

    Tony Parsons Norfolk and Good

    OK, thanks - just curious. Good luck.
  8. That wasn't my experience.
    I remember silk finish being an option, on request, but never mandatory.

    It's true that pictures from a 126 Instamatic probably don't warrant more than 300 ppi scanning. OTOH something like a little Olympus XA is capable of excellent detail rendering. You also have to bear in mind that the resolution of a copy has to be greater than that of the original in order to retain all of the detail that's in the original.

    I honestly don't understand what the problem would be with scanning at 600 ppi. It's the native resolution of many low end flatbed scanners, and won't slow the process down by much, if at all. And as mentioned before, computer storage is cheap and plentiful these days. So that even the average 'laptop' computer won't baulk at processing an 8 Mpx file... or several.
    Just to be clear; I was in no way suggesting using flash memory of any sort for archival storage. I was only giving an example of how little storage is needed for an 8 megapixel file.
    Last edited: Jul 2, 2020
    tomazdrnovsek likes this.
  9. Re-reading your OP, I agree with @rodeo_joe|1 that 300 dpi should be fine for photos of 15 x 10 cm and 600 dpi would be an overkill, just giving you file sizes 4 times as large for no real benefit. A 15 x 10 photo scanned (and cropped to size) at 300 dpi would be about 2 MB. The same photo scanned at 600 dpi would be more than 8 MB.

    The only reasons I can see to use 600 dpi is you scan photos that have physical damage in 'sensitive areas' (faces) or small prints like 6 x 6 cm. The 'optical sharpness' is also worth considering. If film photos are really sharp (or have very small details) then 600 dpi could help to preserve this/these. For photos taken on 'not very expensive cameras' this is perhaps unlikely.
    tomazdrnovsek likes this.
  10. For some photos, it can be useful to use a more precise method than 'eyeballing it' (which is quicker!):
    - add a threshold layer and move the slider from left to right to find the 'blackest points'; mark one of these this with the color sampler tool
    - move the slider from right to left to find the "whitest points"; mark one of these with the color sampler tool too
    - add a new layer above the photo, fill it with 50% grey and set the blending mode to "difference"; go back to the threshold layer and move the slider from left to right to find the 'grey points'; mark one of these with the color sampler tool

    Hide the threshold and grey-filled layers and add a curves layer. In the curves layer, you can set the black, grey and white points based on the color sample points. This should correct the color too.

    tomazdrnovsek likes this.
  11. Yes, the Olympus XA was so good that after losing one on a camping trip, I bought another. But they were mostly owned by advanced amateurs. Some higher-end Instamatics weren't too bad. Most were terrible. My grandmother's 1914 Kodak box camera made better pictures than the typical Instamatic, mainly because the prints were contacts, not enlargements. The real boon for snap shooters was the introduction of compact 35mm cameras with autofocus, built-in electronic flash, and easy loading.

    I agree. If a 4x6-inch snapshot is well composed but unsharp, I'll scan it at 300 dpi, good enough. But most are poorly composed, so I'll scan at 600 dpi and crop it for a better composition, which leaves something like 400-500 dpi.

    BTW, I also tested the optimal resolution for copy prints made at Walgreens. To make a 4x6-inch print, I uploaded the same image at 300 dpi, 600 dpi, and 1200 dpi. The Walgreens prints showed no difference. So now I always scale them to 300 dpi, which is only 1200 x 1800 pixels, or 2.1 megapixels. They upload much faster and still look fine.

    I don't know if the Walgreens machine automatically downsizes everything to 300 dpi or if their printer simply can't resolve more detail. Maybe other places such as Mpix or Shutterfly do better. I recently ordered a 12x12-inch photo book from Shutterfly that includes some low-res pictures made with a smartphone. Some of them are less than 300 dpi. The book hasn't arrived in the mail yet so I don't know how they'll look on paper.
    tomazdrnovsek likes this.
  12. I have a Canon 9950F scanner which lets me scan slides, negatives and prints. I've used it with some success; however, I've had issues with being able to get images with enough quality to enlarge without significant grain. This scanner was last supported by Canon with Windows 7 - not sure about Apple. Fortunately I still have a Windows 7 desktop which I just reconnected to try the scanner again. I'm keeping my fingers crossed that I'm a little smarter and able to get better results this go around.
  13. What film are you using and how is it being processed?
  14. Oh my, its been so long since I used film I don't remember much beyond using Kodachrome, Ektachrome, and sometimes black and white. I sent all of my color film out for processing but processed my own black and white in 1969 & 1970. I used my camera less and less starting in the early 70s through 2001 when I started using an early digital camera that was terrible - press the shutter button and while waiting for the shutter you could go out for dinner and a movie. I still have my old Canon T70 film camera with 4 lenses - 50mm f1.4; 24-36mm f3.5L; 35-105mm, f3.5 macro; 70-210mm f4 macro. I may grab the T70 and ask my granddaughter to go out on a shoot with her Canon F1.

    I love black & white, but could convert color to black and white in post processing. I would appreciate any recommendations that you may have regarding film for my granddaughter and I.
  15. On a much more fundamental level of discussion --

    There is a photographic equivalent of the carpenter's "measure twice, cut once".

    Scanning images is a long and tedious process, so I very much recommend that you
    "Scan once, scan large"
    It takes very little more time to make a large scan than a low resolution scan, and storage is so cheap that filling up a storage device is hard to do.

    You can always downsize without any real problem, but going the other direction sucks.
  16. I think I agree for film, but maybe not for prints.

    Some time ago I was scanning film, and then an 8x10 print with a 3200 dpi scanner.

    The scanner will scan the print at that resolution, but the file is huge (even by today's standard)
    it does take a long time, and I suspect the print doesn't have that much resolution.
  17. That depends on the size of the original film. A contact print from 10"x8" sheet film will obviously have more detail than an 8x enlargement from a 35mm frame.

    In the latter case, no, the print definitely won't contain 3,200 ppi's worth of detail.

    As a rule of thumb, a scan equal to 4,000 ppi from the original negative will garner all the detail that's available. E.g. with a 10x8 blowup from 35mm that would be 4,000/8 = 500 ppi, and even that might be mainly 'empty pixels'.

    Not to mention that a claimed 3,200 ppi 'true optical resolution' is really no such thing when dealing with most flatbed scanners.

    One has to use a little common sense when deciding on a scan resolution.
    Last edited: Jul 21, 2020
  18. I am pretty sure I don't have any 8x10 contact prints.

    It is possible that the scanner gives 27200 pixels in one scan line.
    The chances that those are the right 27200, pretty small.

    First, sampling theory requires delta function sampling at each sample point. Not likely.

    Next, it somehow has to image a line across the glass onto a line CCD sensor with some
    combination of mirrors and lenses. Even if it does manage to get 27200 points, there is very
    likely some distortion on where those points are. This tends to be fairly wide angle, so as
    not to make the scanner too large. One could image a very exact array of lines
    and determine the distortion in the pixel imaging.

    Next, the scan head is moved down the length of the negative or print, with some combination
    of pulleys and gears. To fit within the usual sized box, there is a mirror that moves at half the
    speed of the scanning head, that reflects onto the sensor. There is then a system to compute
    the position of the scan head to, hopefully, close to scanner resolution. (That is, it actually
    does some scanning.)

    In a sort-of best case, it integrates each pixel of 1/3200 inch square, which applies a spatial
    filter to the actual source. Hopefully the source doens't have much more resolution, so we
    don't get any aliasing. Reasonably likely the optics aren't that good, so no problem there.

    In case the desired scan is at less than sensor and scan head movement resolution, it needs
    to downsample the image. That is, low-pass spatial filter to the new spatial frequency,
    then select an appropriate value for each pixel. This can get close to the desired delta
    function, but it is unlikely that they go through the math in both x and y to do that.
    But maybe an approximation to the appropriate low-pass filter and sample point

    For all those reasons, I try not to image 35mm frames on flatbed scanners.

    A dedicated 35mm scanner usually images onto a line sensor with a lens,
    and no mirrors. The distance can be reasonable, so less stretching the optics
    of the lens system. Moving the film the appropriate distance between scans is
    usually easier than moving the mirror/lens system for flatbed scanners.

    I have a cheaper 35mm scanner that images the whole frame at once,
    I suspect using a low-end 2D CCD sensor. Works fine for lower resolution
    needs, such as screen resolution images. Maybe even for 4x6 prints.
    Much faster than moving-film scanners.

    Not to mention that once you have the scan, you have to do something with it that
    might have more imperfections. There are digital optical printers that scan lasers
    across AgBr based paper. Hopefully through lenses that come close to the
    appropriate positions across the paper. As far as I know, 300dpi is usual.
  19. Huh?
    Where does the very specific figure of 27,200 come from? And no tri-line CCD sensor yet made has that many photosites in a single line.
    Also, how can pixels be the 'right' or 'wrong' ones?
    Pixel sampling simply requires a mean brightness or density to be read and digitised from the area being sampled.

    Pixels themselves are dimensionless and have no fixed size.
    I have no idea what the mechanics of a flatbed scanner have to do with selecting a suitable PPI (Pixels Per Inch) figure for scanning.

    Note: Not DPI - Dots Per Inch only applies to half-tone or inkjet printing. The fact that Epson and other scanner makers persistently and perversely use the wrong terminology doesn't make it any more correct.
    Last edited: Jul 22, 2020
  20. 27200 is 8.5*3200, and what Epson claims for the 3200 scanner.
    They also claim interpolated as 12800 pixels/inch, I don't know how they do that.

    The scanner is supposed to sample at equal distances across the source.
    There is a lens between the sensor and source, which might have some distortion,
    so the actual pixels might not be at the desired position. They could be many
    pixels off from the desired position.

    If you look at a small part of the image, this is unlikely to matter, so you still get a high resolution,
    but those pixels could be shifted away from where they are supposed to be. In some very regular
    sources, you might notice.

    From sampling theory, if you sample a frequency limited source at more than twice the highest frequency,
    you can exactly reproduce the original from the samples. If the samples are not at points (delta function),
    but over a wider area, this acts like a low-pass filter on the image. If you don't mind that, that is fine.

    By the way, sampling theory does not require equally space points, though it makes the math a lot
    easier, and also makes reproduction easier. (Besides that sampling theory only works for periodic
    or infinite length sources.)

    Sorry about using the way too common dpi.

Share This Page