• Photo improvers

    From RJH@21:1/5 to All on Sun Nov 3 00:41:14 2024
    I've got a few ropey 100kb or so copies of photos I'd like to print at A4
    size. So something that could lessen the blockiness and make them look more presentable would be useful.

    Anyone any experience of the many 'AI' photo enhancement websites, or desktop software, that does a decent job of such a task?

    --
    Cheers, Rob, Sheffield UK

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Gordon Freeman@21:1/5 to RJH on Mon Nov 4 20:02:27 2024
    RJH <patchmoney@gmx.com> wrote:

    I've got a few ropey 100kb or so copies of photos I'd like to
    print at
    A4 size. So something that could lessen the blockiness and make
    them
    look more presentable would be useful.

    Anyone any experience of the many 'AI' photo enhancement websites,
    or
    desktop software, that does a decent job of such a task?


    The deblocking filters available in some video editors can be very
    successfull but I don't know of an easy way to apply them to still
    images.

    One thing worth trying is the Quantsmooth plugin in Irfanview. If
    you go to Properties > JPG/PCD/GIF > "JPEG Load" and tick "load
    using Quantsmooth" then open the image, it tries to reconstruct the
    lost precision of the quantisation values. The difference on a good
    quality JPEG is marginal and not worth the extra loading time, but
    on a very small or poor quality one it only takes seconds and cleans
    up the artefacts very well, and in fact the author of the plugin
    particularly recommends using when printing low quality images.

    There's also a standalone Quantsmooth progran for Windows or Linux
    downloadable from https://github.com/kilobyte/jpegqs


    I'm sure an AI program would do better, in particular inventing
    plausible detail when enlarging a low res image, but I've no
    experience of using any of them, and whereas Quantsmooth only takes
    a few seconds to run, I understand AI programs typically take hours
    to do their magic.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From RJH@21:1/5 to Gordon Freeman on Wed Nov 6 08:59:08 2024
    On 4 Nov 2024 at 20:02:27 GMT, Gordon Freeman wrote:

    RJH <patchmoney@gmx.com> wrote:

    I've got a few ropey 100kb or so copies of photos I'd like to
    print at
    A4 size. So something that could lessen the blockiness and make
    them
    look more presentable would be useful.

    Anyone any experience of the many 'AI' photo enhancement websites,
    or
    desktop software, that does a decent job of such a task?


    The deblocking filters available in some video editors can be very successfull but I don't know of an easy way to apply them to still
    images.

    One thing worth trying is the Quantsmooth plugin in Irfanview. If
    you go to Properties > JPG/PCD/GIF > "JPEG Load" and tick "load
    using Quantsmooth" then open the image, it tries to reconstruct the
    lost precision of the quantisation values. The difference on a good
    quality JPEG is marginal and not worth the extra loading time, but
    on a very small or poor quality one it only takes seconds and cleans
    up the artefacts very well, and in fact the author of the plugin
    particularly recommends using when printing low quality images.

    There's also a standalone Quantsmooth progran for Windows or Linux downloadable from https://github.com/kilobyte/jpegqs


    I'm sure an AI program would do better, in particular inventing
    plausible detail when enlarging a low res image, but I've no
    experience of using any of them, and whereas Quantsmooth only takes
    a few seconds to run, I understand AI programs typically take hours
    to do their magic.

    Thanks, interesting. I tried the web page version:

    https://ilyakurdyukov.github.io/jpeg-quantsmooth/

    . . . and didn't notice any difference. Think I need something more GUI based
    . . .

    --
    Cheers, Rob, Sheffield UK

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)