Image enlargement software

Discussion in 'Digital Photography' started by Alfred Molon, Nov 19, 2006.

  1. Alfred Molon

    Alfred Molon Guest

    I have some old, lower resolution images which occasionally need to be
    enlarged (pixel count needs to be increased). A friend recommended
    Photozoom, but after giving it a try I'm not that impressed (fail to see
    visible improvements over a simple bicubic interpolation enlargement).
    What other software would be better? I read Roger's enthusiastic posts
    about the Lucy-Richardson stuff - is there a way to try this out?
    Alfred Molon, Nov 19, 2006
    1. Advertisements

  2. Since it's free, start with Irfanview.
    Ed Ruf (REPLY to E-MAIL IN SIG!), Nov 19, 2006
    1. Advertisements

  3. Alfred Molon

    Alfred Molon Guest

    What enlargement tool/method in Irfanview would sou suggest?
    Alfred Molon, Nov 19, 2006
  4. Alfred Molon

    Marvin Guest

    I don't know what improvement you expected to see. Adding
    pixels doesn't improve resolution. I can't for basic
    reasons. All you can really get is interpolated pixels that
    make a smooth image. Some methods do a worse job with a
    particular image.

    Experimenting with different interpolation methods will give
    you some experience. You already noted that two methods
    didn't give much different results. That may. or may not,
    be the case with the next image. I use Paint Shop Pro,
    which offers several interpolation methods that I can apply
    to an image, and choose the one I like best for that image.
    Marvin, Nov 19, 2006
  5. Alfred Molon

    Colin_D Guest


    Try Qimage, free for a month's trial. All printers have a native
    resolution at which they print any file, e.g. 600 ppi for Canons, and
    720 ppi for Epsons. Note that is pixels per inch, the printer dpi (dots
    per inch) is usually much greater, 4800 x 2400 dpi for my Canon, and
    probably similar for Epsons.

    Most image-handling software leaves the ppi conversion to the printer
    software (if you feed an image at, say, 150 ppi to a native 600 ppi
    printer, the printer driver will handle the interpolation) and you are
    then at the mercy of however good that driver is at interpolation.

    Qimage is different in that it is specifically designed to do the
    interpolation before it feeds the image to the printer driver, also it
    utilizes the printer profile, doing a much better job than ordinary
    printer drivers, and the output is noticeably superior.

    Colin D.
    Colin_D, Nov 19, 2006
  6. Another vote for QImage. I've never seen anything better it. The R-L
    (or L-R) stuff is out there but can be a bit non-user friendly, and I
    haven't seen it do better than the QImage pyramid or vector algorithms.

    In Irfanview, try Lanczos. It's a tiny bit better than bicubic, but
    the differences are subtle and sadly, image dependent, which makes this
    stuff all very subjective.. There's no free lunch with enlargement -
    once anything starts guessing pixels, the image content often
    determines how good the result might be!

    Try these for some comparisons - give the second one plenty of time to
    load and read the instructions..
    mark.thomas.7, Nov 20, 2006
  7. Alfred Molon

    Alfred Molon Guest

    Hi Colin, are you perhaps confusing something? I've never heard of any
    printer with 720 ppi. Even the Lightjet printers, which are the best
    currently available on the professional market, print at 300-400 ppi.
    300 ppi is the industry standard.
    By the way, are you sure that your Canon is capable of 4800 dpi? Sounds
    a bit on the high side.
    Alfred Molon, Nov 20, 2006
  8. Alfred,
    Richardson-Lucy is not an interpolation method; it is an
    image restoration (image sharpening) method. I use
    an interpolation method then R-L. In general, I've not
    seen an interpolation algorithm I liked. Every one
    I've used (both commercial image editing and research
    image processing systems) create artifacts. One may
    do well with a certain type of image detail only
    to do poorly on another type. If there were a great
    solution, I would sure like to know about it.

    Roger N. Clark (change username to rnclark), Nov 20, 2006
  9. Photozoom does a reasonably good job on enlarging without creating
    pixelization/jagged edges. Qimage does a very realistic, edge

    I give Qimage my vote, especially for printing, since it also adds a
    workflow. Do note that these programs do not improve resolution, but
    deliver superior (without resampling artifacts) upscaled images.
    ImagesPlus, is the program Roger uses for both its capabilities in
    astronomical imagery and for its implementation of the Richardson-Lucy
    restoration algorithm. There is a time-limited Demo version available
    <> .

    In general, these (RL and similar) restoration methods work best if
    you already have a good model of the lens+sensor unsharpness, a
    so-called Point-Spread-Function (PSF). It might be useful to prepare
    something like a PSF determination and study the website, because
    ImagesPlus may have a bit of a learning curve, and it would allow to
    make the best use of the time offered for testing.
    Bart van der Wolf, Nov 22, 2006
  10. "Roger N. Clark (change username to rnclark)" <>
    wrote in message SNIP
    For understandable reasons, because one doesn't want to enlarge
    sharpening artifacts. However, 'restoration' of up-sampling artifacts
    is not an attractive proposition either.

    I've been experimenting with this as well, and think the best approach
    (although a bit more involved than I state here) is an R-L of the
    original image, followed by an up-sample (e.g. factor 2x like you do,
    which can possibly be characterized and losses restored with a PSF
    based R-L). A good up-sampling method should already reduce artifacts,
    so the second R-L run(s) will only attempt to restore losses, not
    I'm quite pleased with the recent 'Hybrid SE' method from Qimage
    Studio Edition. It preserves edges without stair-stepping, but unlike
    other methods it doesn't create any halo along edges. I know your
    reservations towards Qimage since it produces 8-bit/channel images
    (targeted at printer drivers) and post-processing 8-b/ch images bears
    the risk of posterizing in smooth gradients, like in skies. However,
    already applying R-L at the original size (in 16-b/ch) will reduce the
    amount of substantial sharpening needed after re-sampling (which has a
    large diameter PSF).
    Bart van der Wolf, Nov 22, 2006
  11. Bart,
    Thanks for the info. I'll have to try this too (R-L first).
    I've been working out a new interpolation algorithm for work
    on imaging spectroscopy data. It will be compute intensive
    but will avoid many artifacts I see now.

    Roger N. Clark (change username to rnclark), Nov 22, 2006
  12. Alfred Molon


    Jan 14, 2008
    Likes Received:
    Pretty late on this topic, but I cannot just stand around and not reply. You cand use Reshade to enlarges images without the expected blur associated with standard image interpolation like bicubic.
    The difference from other products like Genuine Fractals or Alien Skin's Blow Up is the greater level of control available. The resulting images can have edge contours ranging from smooth (like blow up) to more realistic ones (like genuine fractals).
    The program also facilitates smoothing or sharpening the texture in the image.

    Hope this helps.
    subpic, Jan 14, 2008
    1. Advertisements

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.