Upscaling Programs - Qimage, Genuine Fractals, pxl SmartScale

Discussion in 'Digital Photography' started by Hans-Georg Michna, Oct 25, 2003.

  1. Has anybody tried these programs and can compare them?

    Qimage, Genuine Fractals, pxl SmartScale (by Extensis)

    It seems they can work magic by putting matching information
    into pictures to raise the resolution without creating an
    unsharp picture.

    I've been looking for a program that keeps edges sharp while
    upscaling, but perhaps fractals also solve this problem somehow.
    I'd like to see some sample details. Does anybody know a web
    site like that?

    Hans-Georg

    --
    No mail, please.
    Hans-Georg Michna, Oct 25, 2003
    #1
    1. Advertising

  2. Hans-Georg Michna

    Rafe B. Guest

    On Sat, 25 Oct 2003 06:31:01 +0200, Hans-Georg Michna
    <> wrote:

    >Has anybody tried these programs and can compare them?
    >
    >Qimage, Genuine Fractals, pxl SmartScale (by Extensis)
    >
    >It seems they can work magic by putting matching information
    >into pictures to raise the resolution without creating an
    >unsharp picture.
    >
    >I've been looking for a program that keeps edges sharp while
    >upscaling, but perhaps fractals also solve this problem somehow.
    >I'd like to see some sample details. Does anybody know a web
    >site like that?



    You can try Genuine Fractals for yourself; there's a free
    download.

    Personally I regard each of these products as an exercise
    in futility and delusion.

    Computers are wonderful at manipulating and processing
    data. But image data can't be manufactured from thin air.



    rafe b.
    http://www.terrapinphoto.com
    Rafe B., Oct 25, 2003
    #2
    1. Advertising

  3. Rafe B. <> wrote:

    >You can try Genuine Fractals for yourself; there's a free
    >download.
    >
    >Personally I regard each of these products as an exercise
    >in futility and delusion.
    >
    >Computers are wonderful at manipulating and processing
    >data. But image data can't be manufactured from thin air.


    Rafe,

    thanks for the hint.

    As to the futility of upscaling, you're probably mistaken here,
    for several reasons. One is that the pixel detail in many photos
    does not matter, as long as the structures, like edges, remain
    sharp when you scale up. Another is that you often have
    structures like leaf cover where again the exact detail does not
    matter at all, as long as the structure looks natural. When you
    take a photo of a forest, it simply does not matter whether the
    tiny leaves of a few dozen pixels have the correct shape, as
    long as they have any shape at all. Or take a photo of a bearded
    person. Does it matter what microshape the edge of the beard
    has? No, it doesn't. This is where fractals come into play.

    If you're really interested, have a look at the work of a
    Faroudja image processor chip while displaying a DVD movie at
    high resolution. This is something hard to believe, until you
    watch it consciously, particularly the edge smoothing
    (antialiasing).

    And a more futuristic reason is that programs are conceivable
    that fill in the missing details using knowledge about the real
    world. For example, such a program could know what a human face
    and eye looks like and could fill in the missing pixels
    accordingly, for example for small faces in the background. We
    don't have that yet, but I'm pretty sure we will one day. Of
    course such programs can go wrong occasionally, but overall they
    will be very much more useful than harmful.

    Hans-Georg

    --
    No mail, please.
    Hans-Georg Michna, Oct 25, 2003
    #3
  4. Hans-Georg Michna

    Rafe B. Guest

    On Sat, 25 Oct 2003 08:00:24 +0200, Hans-Georg Michna
    <> wrote:

    >Rafe B. <> wrote:
    >
    >>You can try Genuine Fractals for yourself; there's a free
    >>download.
    >>
    >>Personally I regard each of these products as an exercise
    >>in futility and delusion.
    >>
    >>Computers are wonderful at manipulating and processing
    >>data. But image data can't be manufactured from thin air.

    >
    >Rafe,
    >
    >thanks for the hint.
    >
    >As to the futility of upscaling, you're probably mistaken here,
    >for several reasons. One is that the pixel detail in many photos
    >does not matter, as long as the structures, like edges, remain
    >sharp when you scale up. Another is that you often have
    >structures like leaf cover where again the exact detail does not
    >matter at all, as long as the structure looks natural. When you
    >take a photo of a forest, it simply does not matter whether the
    >tiny leaves of a few dozen pixels have the correct shape, as
    >long as they have any shape at all. Or take a photo of a bearded
    >person. Does it matter what microshape the edge of the beard
    >has? No, it doesn't. This is where fractals come into play.




    All I know is that I gave GF a try a couple years back,
    compared it against plain old bicubic resampling and
    found it utterly lame.

    In fact what I found was that GF added lots of artifacts
    to my images which didn't help the images at all. In
    short, I concluded that GF was nothing more than yet-
    another lossy compression algorithm, and not even
    as good as high-quality JPG.

    Upsampling is something I do only when there is
    no other choice.

    PS: I do understand a bit about the nature of fractals,
    and that's what led to my initial interest in the product.
    But having tried it, I consider the name to be pure
    marketing hype, in the case of the product known
    as "Genuine Fractals."

    Some foks think it's the cat's meow. My own
    experience says, don't bother.


    rafe b.
    http://www.terrapinphoto.com
    Rafe B., Oct 25, 2003
    #4
  5. Hans-Georg Michna

    HRosita Guest

    >Rafe B. wrote:

    >All I know is that I gave GF a try a couple years back,
    >compared it against plain old bicubic resampling and
    >found it utterly lame.
    >
    >In fact what I found was that GF added lots of artifacts
    >to my images which didn't help the images at all. In
    >short, I concluded that GF was nothing more than yet-
    >another lossy compression algorithm, and not even
    >as good as high-quality JPG.
    >


    Genuine Fractals, like other software has changed considerably from the initial
    product. It works best with larger files and I find it quite good in enlarging
    slide scans especially and even 5 MP digital images.

    I prefer it to bicubic resampling.
    I will try to post cropped samples this weekend.



    Rosita
    HRosita, Oct 25, 2003
    #5
  6. "Hans-Georg Michna" <> wrote in message
    news:...
    > Has anybody tried these programs and can compare them?
    >
    > Qimage, Genuine Fractals, pxl SmartScale (by Extensis)


    I've tried several and my findings are similar to
    http://www.abdn.ac.uk/diss/ltu/pmarston/photography/Tests/scaling.html .

    > It seems they can work magic by putting matching information
    > into pictures to raise the resolution without creating an
    > unsharp picture.


    They all create the impression of resolution by accentuating sharp edge
    contrast more than soft edge contrast.

    > I've been looking for a program that keeps edges sharp while
    > upscaling, but perhaps fractals also solve this problem somehow.
    > I'd like to see some sample details. Does anybody know a web
    > site like that?


    See above.

    Also important is to consider why you rescale.
    The most common reason is preparing for larger size prints than the file
    size would normally allow. With this in mind, I find Qimage superior to all
    the above mentioned products, and it econimizes on paper use if you print
    several images.

    If you realize that most printer drivers internally resize by a factor of
    about 4, on top of the resizing you do to reach a decent size, it's not a
    mystery why Qimage output looks soo good. It does all resizing in one go, it
    uses superior interpolation techniques to choose from, it enhances sharp
    edge detail, and it sharpens luminance *after* resizing. All this without
    altering your original file.

    As a bonus, Qimages allows to print to file so you can prepare output for,
    say, a Frontier printer. It can also be (mis-)used for preparing display on
    a monitor, and I've tried that on the (JPEG) example given in the webpage
    mentioned above. To me, the overall result beats all of the other examples
    when looked at on screen. It avoids the plastic look and retains the
    original pixel detail (cloth detail versus metal and leather) best. Printing
    them gives Qimage even more of an edge.

    Bart
    Bart van der Wolf, Oct 25, 2003
    #6
  7. "Rafe B." <> wrote in message
    news:...
    SNIP
    > Upsampling is something I do only when there is
    > no other choice.
    >
    > PS: I do understand a bit about the nature of fractals,
    > and that's what led to my initial interest in the product.
    > But having tried it, I consider the name to be pure
    > marketing hype, in the case of the product known
    > as "Genuine Fractals."


    I fully agree. In theory fractal imaging is a promising concept, but what GF
    does is posterize into sharp edged, reduced feature areas.
    Also important to realize is that GF recommends to start with a high
    resolution image from the beginning. Kind of defeats the need...

    > Some foks think it's the cat's meow. My own
    > experience says, don't bother.


    It can work on particular subjects with many sharp edges, e.g. images of
    some architecture, viewed from a distance.

    For most other subjects, there are better alternatives.

    Bart
    Bart van der Wolf, Oct 25, 2003
    #7
  8. Hans-Georg Michna

    Bill Hilton Guest

    >From: Hans-Georg Michna

    >Has anybody tried these programs and can compare them?
    >
    >Qimage, Genuine Fractals, pxl SmartScale (by Extensis)
    > ... Does anybody know a web site like that?


    This site tries several different sharpening techniques, including GF, Lanczos
    (Qimage), S-Spline, regular Photoshop bi-cubic and "stair interpolation",
    basically bicubic in 10% increments. http://www.fredmiranda.com/SI/index.html
    Keep in mind he's selling a Photoshop action, but SI does very well in this
    test.

    >It seems they can work magic by putting matching information
    >into pictures to raise the resolution without creating an
    >unsharp picture.


    As others have pointed out there is no free lunch or we'd all be shooting 2
    Mpix digital cameras and upsampling. But sometimes you have to upsample an
    image and as you'd expect some of the products do a better job than others,
    though to some extent it's image structure based.

    I recently got an order for a 20x30" print from an image which had been shot on
    35 mm (would normally shoot on 6x7 cm and have no problem with this). The
    place I use for large prints gives a big discount if you send in the file at
    the exact size their machine wants, in this case 304.8 ppi, so I had to
    interpolate. I printed off 1/4 th of the image at 10x15" to judge sharpness
    and interpolated using GF (have a free copy bundled with my scanner), regular
    bicubic, Stair Interpolation at 10% and a couple of other ways as well and
    found SI did the best job, though not by a whole lot.

    Supposedly GF works best on very large blowups and I've seen tests on vector
    art where GF did better than bicubic, but for photos with the size enlargement
    I needed SI was a better solution. Given the price of GF I think it's vastly
    overrated for photographers. You can create your own SI action easily in
    Photoshop so it's essentially free.

    Bill
    Bill Hilton, Oct 25, 2003
    #8
  9. Hans-Georg Michna

     Guest

    I moderately upsample most of my digital images. I have a 2MP camera I
    won't part with (C2100UZ) because of the fabulous zoom, IS, and sharp
    lens. I usually upsample to 3MP for printing. I have used Extensis
    (free download trial), Fred Miranda's Stair Interpolation in Photoshop
    6, plain old bicubic in Photoshop, and Qimage. My preference is Qimage.
    Qimage now uses a method they call Vector. This is an improvement over
    their prior best method, Lanczos, which is in some of those comparison
    sites.

    I have done all the side by sides using several of my own images (use
    copies, not originals please). I just feel the Qimage with either
    Lanczos or now Vector gives the most pleasing results.

    I have even taken a sharp 2MP image, scaled it up to about 8MP with
    Qimage Vector and cropped out a small area, about 25% of the image.
    Then, with a touch up in Neat Image to remove some noise, it makes an
    excellent 4X6. Almost as though I had originally zoomed into that area
    optically. I also did a crop, about 2/3 of a 2MP image upsampled to
    8MP, that I printed as an 8X10. I think it looks fine.

    Maybe one day when I feel the need to upgrade to a 6MP camera, I won't
    need upsampling, but for now my pics are very good for me. I would
    recommend Qimage.

    BTW, I also think Qimage does a better job at level adjustment and color
    correction than I can get myself in Photoshop. Maybe someone more
    expert in Photoshop could do better, but for me Qimage is better.





    In article <>, hans-
    says...
    > Has anybody tried these programs and can compare them?
    >
    > Qimage, Genuine Fractals, pxl SmartScale (by Extensis)
    >
    > It seems they can work magic by putting matching information
    > into pictures to raise the resolution without creating an
    > unsharp picture.
    >
    > I've been looking for a program that keeps edges sharp while
    > upscaling, but perhaps fractals also solve this problem somehow.
    > I'd like to see some sample details. Does anybody know a web
    > site like that?
    >
    > Hans-Georg
    >
    >
    , Oct 25, 2003
    #9
  10. Hans-Georg Michna, Oct 25, 2003
    #10
  11. X-Original-Message-ID:
    <>
    X-Agent-Group: rec.photo.digital
    X-Agent-Format: 2 1 0 0 1 56700 1 0 1 1 "*" 0
    X-Intro: "
    <> wrote:\n"

    Thanks to all for the interesting information!

    It seems that Qimage is liked by several people here. Other
    people say things like this (in other channels, anonymized):

    "Qimage uses bicubic as one of it's choices although the best
    seems to be their vector method. There's no comparison in the
    results, GF and pxl SS both do about the same quality job and
    blow away what Qimage can do."

    What I have seen on the one sample web page
    http://www.abdn.ac.uk/diss/ltu/pmarston/photography/Tests/scaling.html
    seems to confirm this, but one sample may not be the whole
    truth.

    I guess it also depends on what you want to do. If you want to
    upscale a 4 Megappixel picture to 6 Megapixels, Qimage may well
    be the best. If you want to upscale a 2 Megapixel picture to 8
    Megapixels for printing, you may need something else, like pxl
    SmartScale.

    Hans-Georg

    --
    No mail, please.
    Hans-Georg Michna, Oct 25, 2003
    #11
  12. Hans-Georg Michna

    Rick Guest

    "Hans-Georg Michna" <> wrote in message news:...
    > "Bart van der Wolf" <> wrote:
    >
    > >I've tried several and my findings are similar to
    > >http://www.abdn.ac.uk/diss/ltu/pmarston/photography/Tests/scaling.html .

    >
    > Bart,
    >
    > pxl SmartScale is it. That's the only sample that was really
    > sharp. All others are fuzzy.


    That's not "sharp", that's posturized. The edges aren't at all
    natural looking. Of those sample images GF and Fred's SI
    are a close tie.

    Rick
    Rick, Oct 25, 2003
    #12
  13. Hans-Georg Michna

    someone Guest

    On Sat, 25 Oct 2003 14:25:45 GMT,
    <> posted:

    >I moderately upsample most of my digital images. I have a 2MP camera I
    >won't part with (C2100UZ) because of the fabulous zoom, IS, and sharp
    >lens. I usually upsample to 3MP for printing. I have used Extensis
    >(free download trial), Fred Miranda's Stair Interpolation in Photoshop
    >6, plain old bicubic in Photoshop, and Qimage. My preference is Qimage.
    >Qimage now uses a method they call Vector. This is an improvement over
    >their prior best method, Lanczos, which is in some of those comparison
    >sites.
    >
    >I have done all the side by sides using several of my own images (use
    >copies, not originals please). I just feel the Qimage with either
    >Lanczos or now Vector gives the most pleasing results.
    >
    >I have even taken a sharp 2MP image, scaled it up to about 8MP with
    >Qimage Vector and cropped out a small area, about 25% of the image.
    >Then, with a touch up in Neat Image to remove some noise, it makes an
    >excellent 4X6. Almost as though I had originally zoomed into that area
    >optically. I also did a crop, about 2/3 of a 2MP image upsampled to
    >8MP, that I printed as an 8X10. I think it looks fine.
    >
    >Maybe one day when I feel the need to upgrade to a 6MP camera, I won't
    >need upsampling, but for now my pics are very good for me. I would
    >recommend Qimage.
    >
    >BTW, I also think Qimage does a better job at level adjustment and color
    >correction than I can get myself in Photoshop. Maybe someone more
    >expert in Photoshop could do better, but for me Qimage is better.


    so you don't use photoshop at all?
    someone, Oct 26, 2003
    #13
  14. "Rick" <> wrote:

    >"Hans-Georg Michna" <> wrote in message news:...


    >> "Bart van der Wolf" <> wrote:


    >> >I've tried several and my findings are similar to
    >> >http://www.abdn.ac.uk/diss/ltu/pmarston/photography/Tests/scaling.html .


    >> pxl SmartScale is it. That's the only sample that was really
    >> sharp. All others are fuzzy.


    >That's not "sharp", that's posturized. The edges aren't at all
    >natural looking. Of those sample images GF and Fred's SI
    >are a close tie.


    Rick,

    I guess you have to look at the real thing from time to time to
    remember what an edge looks like in nature. :)

    Hans-Georg

    --
    No mail, please.
    Hans-Georg Michna, Oct 26, 2003
    #14
  15. "Hans-Georg Michna" <> wrote in message
    news:...
    SNIP
    > >> pxl SmartScale is it. That's the only sample that was really
    > >> sharp. All others are fuzzy.

    >
    > >That's not "sharp", that's posturized. The edges aren't at all
    > >natural looking. Of those sample images GF and Fred's SI
    > >are a close tie.


    I tend to agree with that. Look at the metal buttons on the sporran (pouch).
    The Extensis plug-in completely distroys any sense of 3-dimensionality. The
    white shirt looks like it could have been made from plastic, leather, or
    cloth. The belt buckle looks like a fresco instead of a shiny metal The
    other methods retain a more faithful redition of the original.

    > I guess you have to look at the real thing from time to time to
    > remember what an edge looks like in nature. :)


    As said before, edges are important for the impression of sharpness, but the
    enlargements shown are only 4x enlargements and so much information is
    already lost. Try something like this on a human face, and skin will look
    like it's a wax-statue instead of a real warm person. Texture is also very
    important, and the extensis plug-in IMHO destroys more than it adds.

    The only use I see for it, is when the viewing distance is so large that the
    texture will be impossible to resolve by eye. Only then the edge sharpness
    helps to make out shapes, and give an impression of sharpness.

    Bart
    Bart van der Wolf, Oct 26, 2003
    #15
  16. Hans-Georg Michna

     Guest

    In article <>,
    says...
    >
    > On Sat, 25 Oct 2003 14:25:45 GMT,
    > <> posted:
    >
    > >I moderately upsample most of my digital images. I have a 2MP camera I
    > >won't part with (C2100UZ) because of the fabulous zoom, IS, and sharp
    > >lens. I usually upsample to 3MP for printing. I have used Extensis
    > >(free download trial), Fred Miranda's Stair Interpolation in Photoshop
    > >6, plain old bicubic in Photoshop, and Qimage. My preference is Qimage.
    > >Qimage now uses a method they call Vector. This is an improvement over
    > >their prior best method, Lanczos, which is in some of those comparison
    > >sites.
    > >
    > >I have done all the side by sides using several of my own images (use
    > >copies, not originals please). I just feel the Qimage with either
    > >Lanczos or now Vector gives the most pleasing results.
    > >
    > >I have even taken a sharp 2MP image, scaled it up to about 8MP with
    > >Qimage Vector and cropped out a small area, about 25% of the image.
    > >Then, with a touch up in Neat Image to remove some noise, it makes an
    > >excellent 4X6. Almost as though I had originally zoomed into that area
    > >optically. I also did a crop, about 2/3 of a 2MP image upsampled to
    > >8MP, that I printed as an 8X10. I think it looks fine.
    > >
    > >Maybe one day when I feel the need to upgrade to a 6MP camera, I won't
    > >need upsampling, but for now my pics are very good for me. I would
    > >recommend Qimage.
    > >
    > >BTW, I also think Qimage does a better job at level adjustment and color
    > >correction than I can get myself in Photoshop. Maybe someone more
    > >expert in Photoshop could do better, but for me Qimage is better.

    >
    > so you don't use photoshop at all?
    >
    >


    If I have a gem of a shot (a perfectly cute expression from one of my
    kids for instance) that is maybe a little blurry due to motion or a bad
    focus, I will use Photoshop to try to rescue enough to print. Some of
    the sharpening schemes are useful to me. Also, if I have something
    where I want to bring out shadow details, I use Fred Miranda's Shadow
    Recovery in Photoshop and it works well.

    My sequence is usually:
    1) Photoshop if I need to address any of the above.
    2) Qimage for upsample and color/exposure correction and cropping if
    needed.
    3) Neat Image to take out any noise (I never use 100%, usually 70%).
    , Oct 26, 2003
    #16
  17. "Bart van der Wolf" <> wrote:

    >"Hans-Georg Michna" <> wrote in message
    >news:...


    >> I guess you have to look at the real thing from time to time to
    >> remember what an edge looks like in nature. :)


    >As said before, edges are important for the impression of sharpness, but the
    >enlargements shown are only 4x enlargements and so much information is
    >already lost. Try something like this on a human face, and skin will look
    >like it's a wax-statue instead of a real warm person. Texture is also very
    >important, and the extensis plug-in IMHO destroys more than it adds.


    Bart,

    what I saw was that it only made high-contrast edges sharp, so
    it wouldn't do anything like that to texture, only to real
    edges.

    But of course I can't be sure from just one sample.

    Hans-Georg

    --
    No mail, please.
    Hans-Georg Michna, Oct 26, 2003
    #17
    1. Advertising

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

It takes just 2 minutes to sign up (and it's free!). Just click the sign up button to choose a username and then you can ask your own questions on the forum.
Similar Threads
  1. speedburn

    genuine fractals ser no req'd thanks

    speedburn, Jul 23, 2003, in forum: Digital Photography
    Replies:
    3
    Views:
    373
    Webkatz
    Jul 25, 2003
  2. Click

    Genuine Fractals questions

    Click, Jul 27, 2003, in forum: Digital Photography
    Replies:
    2
    Views:
    445
    Bill Hilton
    Jul 27, 2003
  3. Re: Genuine fractals trial

    , Jul 31, 2003, in forum: Digital Photography
    Replies:
    4
    Views:
    530
    Bart van der Wolf
    Aug 2, 2003
  4. Lisa Horton

    16x20 prints from 10D, Genuine Fractals?

    Lisa Horton, Sep 3, 2003, in forum: Digital Photography
    Replies:
    5
    Views:
    665
    Bill Hilton
    Sep 4, 2003
  5. Brute Force
    Replies:
    1
    Views:
    593
    Hans-Georg Michna
    Feb 13, 2004
Loading...

Share This Page