Working in Photoshop at maximum image size/resolution

Discussion in 'Digital Darkroom' started by robert goldstein, Jul 7, 2006.

  1. I routinely scan 35mm slides at the maximum resolution of my scanner (5400ppi),
    so as to get as much detail off the slide as possible. However, in 16bit mode,
    this creates unwieldy files that are about 215Mb in size. After editing in
    Photoshop with multiple layers, the files can easily exceed 1Gb in size.

    What is the best way to deal with such huge files? Is it always best to edit
    at maximum size and then resample/resize for final output, or is resizing
    before editing acceptable? The latter approach could be problematic if
    multiple output sizes are required, either at the time of initial scanning or
    in the future.

    I guess what I am seeking is a proper workflow for handling and storing my
    scans. Any advice will be appreciated.
  2. Assuming that your scanner is actually scanning at that resolution (not interpolating) it might simply be overdoing it.... You should be scanning at the highest "native" resolution and let Photoshop deal with resizing algorithms.

    What really matters is the final size for the desired print - if you goal is to print 4"x6" only 1200x1800 pixels is required... if you want to be able to print a 16x20" you will still "only" need 4800x6000 pixels.... any more is overkill.
  3. My question is not about how to resize the image, but rather what is the best resolution at which to edit it. Sometimes, I do not even know the ultimate output size when I begin working on an image
  4. Use the largest print size you may need... if you have never printed anything bigger than a 16x20 what is the point of using a resolution for a 30x40 poster? Beyond a certain point you will also reach the resolution limits of your slides anyway.

    Hope this helps
  5. Robert,

    I wondered the same as Denis whether you were scanning at the scanner's interpolated, or native (aka 'optical') resolution. The odds are that the file will be smaller if scanned to the optical resolution, which goes towards your question of "What is the best way to deal with such huge files?" Query: are they bigger than they need to be in order to capture "as much detail off the slide as possible"?

    If you've scanned for the highest capture of detail, it seems the logic of the workflow is to produce a file in the same vein by editing "at maximum size", saving as a 16bit tif or psd except for the final sharpening, and making that the file that gets resampled or downsized with a final sharpening (and maybe other edits, too, for the specific output).

    In my workflow I end up with archiving the original scan 16bit tif, the edited 16bit psd (with all layers), copies of the final output(s), and a text file of notes.

    Don E
  6. If you absolutely feel the need for large files, because of your printing output sizes, the only other method that I believe is non-destructive editting, besides layers, is the History Brush Tool in combination with it's Blending Mode Options. What it allows you to do is slectively brush in different blending modes to different areas of the image with a brush...kinda paint it in. For instance the blending mode called Multiply can darken certain areas by brushing it in. Russell Brown (go about 1/3 of the way down the page) has a tutorial called History 101 on this. The part on using JUST the History Brush is maybe half way thru the need Quicktime loaded for these (it's free). What's nice is the file size remains the same thru out some pretty massive editting changes.
  7. I have this problem fairly routinely, as I've been printing on Super A/B3 and larger for the last few months. Roughly, this is my workflow:

    1. Scan at a resolution slightly higher than needed.
    2. Retouch dust and scratches
    3. Save file as a master copy, then reduce the file size to something managable, say 30 MB (8-bit, so that I lose no functionality). Save this as my working copy.
    4. Perform all further edits on the working copy, using adjustment layers and layer masks, primarily curves. Save all important paths as alpha channels.
    5. When finished, resize the working copy back to the original size, and then copy all the adjustment layers over to the master. If I needed to work with a raster layer (eg, a layer that consists of pixels rather than mathematical adjustments - is there a better term for that) then I recreate those with the master file's original data, and copy the appropriate masks.

    This allows me to work at the pace of a small file, while still retaining all the original detail. All those layer masks are essentially single-channel data that don't need to be 16-bit, and all the adjustment layers are really just math performed on the pixels, so when I copy them over, they work just as well. Printers are 8-bit anyway, so the whole reason we're using 16-bit is just to get better rounding errors.
  8. I do all my level, curve, hues etc. with adjustment layers on a 1/4 size file (that is, 1/2 the
    linear dimensions) When I'm done I have a script that blows it back up to full size, swaps
    back the full-rez image, and saves it out as a flattened 8-bit file to do any retouching,
    spotting, etc on. Reflating takes a while, so I generally do this on a batch basis--edit a
    bunch of pics, and have the computer reflate them while I'm having dinner or something.

    Apparently there's something called Smart Objects that can make this even simpler in CS2,
    but I'm still stuck in CS1.
  9. Robert, think about 3-4 Gb of memory and how many of your problems it will solve. I have 4 Gb and never have a problem even with large MF scans with many layers.

    Short of buying memory, try immediate downsampling of the scan TIF to the largest file size you think there is a 98% probability you may probably use. The file will get reasonable size and easy to work with. When you hit that other 2% that you want to print huge, then try all of the previous suggested alternatives.

    Bottom line: don't work up every file so you can print huge just because some day you might want to do one or two that way. Work for the majority and deal with the exceptions when they occur.
  10. I also scan 5400dpi images of films at 16-bit, and get 250~MB files.

    I run my editing off a Athlon @2.34GHz and 1.5GB RAM. I dedicate 50% of this RAM to Photoshop, so that PS can keep 2-3 edits of an image before using slow hard disk swap files. I imagine with 1.5GB you should be able to safely dedicate 75% or so and allow for 1GB of revisions to be done in memory. Compared to your film scanner, RAM is pretty cheap, so invest in it, it'll save you frustration!
  11. I appreciate all of the input so far. Keep it coming.

    Bob Michaels, I have 1.8Gb RAM installed, but is there not a limit to the RAM that
    Photoshop CS can actually utilize?
  12. Robert, GS2 can use 3 Gb once you set the 3 Gb switch (a software step in XP). Having 4 Gb leaves 1 Gb for everything else.
  13. Robert, I just noticed you said CS, not CS2. CS can only use 2 Gb. It takes CS2 for 3 Gb. So I'd try 3 Gb of memory and allocate 2 Gb to Photoshop if you have CS.

    Fortunately memory is cheap now.
  14. First, save your original raw scan and work on a copy.
    Second, when your copy with layers gets too large, make a flattened copy and continue working on this version, until it too gets too large and then make another flattened copy etc.

    For example, on my initial copy of my raw scan, I may run Neat Image and PK Capture Sharpener, then make a flattened copy and perform local contrast enhancement, curves etc. If I need to do any creative sharpening, I do it on a flattened copy. If done correctly, you never need to work on extremely large files and you preserve your ability to tweak your edits as you would if everything was in one file (although not quite as easily).
  15. jtk


    IMO Minolta's "5400" detail resolution is the same as its 4000ppi, but results in a much larger file (4000 is about 120Mb). A Japanese test reported this about this a year or two ago. My own comparison of 5400 Vs Nikon's 4000 showed them equal, both rendering grain very sharply. Certainly, 4000ppi is more than enough for grain sharp prints to my limit of 12X18...waaay better than an optical enlarger could accomplish.

    Using 4000ppi would sharply reduce file size, not to mention time.
  16. You may want to read this thread:

    Note Erik's recommendation about scan resolution. "I would however scan at 5400 dpi (16 bit) and if needed downsample in Photoshop. Downsampling is what Minolta's SW is doing too, but it does a less good job than Photoshop, beause it makes the scan head move with bigger steps in one direction and downsamples the resolution of the line-CCD in the other direction. Because of this scanning method actually less input information is used to calculate the downsampled information from."

Share This Page