Does bicubic resampling make a noticeable improvement to printed photos?

Discussion in 'Digital Photography' started by Lee, Apr 29, 2006.

  1. Lee

    Lee Guest

    Has anyone out there actually compared the printed results of using
    photoshop bicubic resampling to "interpolate up" the pixels in order to
    print a larger photo (example: 20x30) than your digital camera would
    normally be able to support? Were the results better with resampling
    than just turning it off and sending the photo to the printer with
    whatever the resolution drops to when you print it at such a large size
    (example: 8.2megapixel camera 3456x2304pixels. According to Photoshop
    this image has a resolution of 115 ppi at 20x30. Printers generally
    recommend 300 dpi minimum). Should I resample this image to
    300ppi?Thanks
    Lee, Apr 29, 2006
    #1
    1. Advertising

  2. "Lee" <> wrote in message
    news:...
    > Has anyone out there actually compared the printed results of using
    > photoshop bicubic resampling to "interpolate up" the pixels in order to
    > print a larger photo (example: 20x30) than your digital camera would
    > normally be able to support? Were the results better with resampling
    > than just turning it off and sending the photo to the printer with
    > whatever the resolution drops to when you print it at such a large size
    > (example: 8.2megapixel camera 3456x2304pixels. According to Photoshop
    > this image has a resolution of 115 ppi at 20x30. Printers generally
    > recommend 300 dpi minimum). Should I resample this image to
    > 300ppi?Thanks


    I've seen several "trials" in the photographic press, and I've been hard
    pressed to spot the difference between the various techniques, even when the
    writer pointed out the "obvious". I guess it might be more obvious if
    printed on decent paper.
    Don't you use these resampling things to raise the file sizes so that your
    low res image isn't rejected automatically by Alamy etc.?

    --
    M Stewart
    Milton Keynes, UK
    http://www.megalith.freeserve.co.uk/oddimage.htm
    Malcolm Stewart, Apr 29, 2006
    #2
    1. Advertising

  3. Lee

    Stacey Guest

    Lee wrote:

    > Has anyone out there actually compared the printed results of using
    > photoshop bicubic resampling to "interpolate up" the pixels in order to
    > print a larger photo (example: 20x30) than your digital camera would
    > normally be able to support?


    The printer software upsamples so the question is does the printer software
    do a better job than photoshop or something like Qimage does? The answer is
    no. Qimage upsamples every image to the printer native rez (and then
    resharpens before feeding this file to the printer), which on my canon
    i9900 is 600DPI instead of allowing the printer software to do this. It's
    not a HUGE improvement buit it's visible,
    --

    Stacey
    Stacey, Apr 30, 2006
    #3
  4. Lee

    bmoag Guest

    Inkjet photo printers print natively at 300dpi.
    Printer manufacturers may cite specs ranging from slightly less to
    significantly higher but the fundamental number is 300dpi for the highest
    quality ink jet photo printers.
    You can set the print DPI smaller: many if not most images printed at 250
    dpi are not distinguishable from images printed at 300 dpi.
    A Canon 9900 does not print at anything approaching 600dpi as this is
    physically impossible given droplet size and the interaction of ink and
    paper. Some laser printers can indeed print tiny text at greater than 300
    dpi but you need a magnifying glass to see it.
    When you work on your masterpiece in Photoshop if you check image size you
    can see a wide variety of DPI settings. For example a Nikon D70 RAW image in
    8 bit color will open at around 250 dpi. This reflects the great wisdom of
    Nikon's dSLR engineers with regard to optimal settings for their cameras who
    have made decisions intending to keep users out of trouble.
    When you have completed all your futzing with your image and want to print
    it you can either let the printer driver use its immutable algorithms to
    convert that image to the appropriate size and dpi or you can perform the
    conversion yourself. If the image and your manipulations of the image are of
    reasonably high quality, and you understand the process of color managed
    printing, and rarely print greater than 8.5x12 my experience is that how the
    conversion is made, by you or the printer algorithm, matters not.
    That is: if you make the conversion in Photohop or merely use print preview
    and pull the image to whatever size you want on the print the result is more
    often than not the same if you begin with a good quality image and
    understand color management.
    CS2 in particular gives you several options with regard to how to convert
    your masterpiece to the appropriate size and dpi for printing. The most
    intriguing is "bicubic sharper", which is not the default algorithm. This
    can indeed make a visible difference when making large prints out of less
    than stellar originals, particularly jpegs from P&S cameras, but it is not
    necessarily an overwhelming improvement over the default algorithm.
    The only way to get a handle on how these things affect your printing is to
    experiment.
    bmoag, Apr 30, 2006
    #4
  5. Lee

    Stacey Guest

    bmoag wrote:

    > Inkjet photo printers print natively at 300dpi.


    Wrong. Canon native hardware is 600, epson is 720. What do you think those
    1200 and 1440 number are about?

    > Printer manufacturers may cite specs ranging from slightly less to
    > significantly higher but the fundamental number is 300dpi for the highest
    > quality ink jet photo printers.


    What you're reading is people saying 300DPI is what you need to start with
    before printer driver upsampling to get a good print.

    > You can set the print DPI smaller: many if not most images printed at 250
    > dpi are not distinguishable from images printed at 300 dpi.


    Exactly because the printer driver upsamples it to the native hardware rez
    before printing.


    > A Canon 9900 does not print at anything approaching 600dpi as this is
    > physically impossible given droplet size and the interaction of ink and
    > paper. Some laser printers can indeed print tiny text at greater than 300
    > dpi but you need a magnifying glass to see it.


    You won't see the over 300DPI starting point with most images difference but
    that 600DPI is the size file the printer is looking for. Just because you
    need a magnifying glass to see it doesn't change anything.

    Given people bitch about sensor noise that you can only see at 100% on a
    screen and can't be seen in a print, seems odd to discount somthing you can
    actually see if you look at the print closely.

    > When you have completed all your futzing with your image and want to print
    > it you can either let the printer driver use its immutable algorithms to
    > convert that image to the appropriate size and dpi or you can perform the
    > conversion yourself.


    Which is what I just said, you just are using the wrong information as to
    what the printer hardware is looking for.

    > If the image and your manipulations of the image are
    > of reasonably high quality, and you understand the process of color
    > managed printing, and rarely print greater than 8.5x12 my experience is
    > that how the conversion is made, by you or the printer algorithm, matters
    > not.


    So you've tried Qimage then?

    > That is: if you make the conversion in Photohop or merely use print
    > preview and pull the image to whatever size you want on the print the
    > result is more often than not the same if you begin with a good quality
    > image and understand color management.


    But you still have ignored that the final sharpening should be done before
    the printer takes over. If you send the file to the printer at under it's
    NATIVE hardware rez, the printer driver is going to upsample it as the last
    step, not ideal. It's not a big difference but it's there if you critically
    look at the final prints.

    --

    Stacey
    Stacey, May 1, 2006
    #5
  6. Lee

    2 Guest

    "Stacey" <> wrote in message
    news:...
    > bmoag wrote:
    >
    >> Inkjet photo printers print natively at 300dpi.

    >
    > Wrong. Canon native hardware is 600, epson is 720. What do you think those
    > 1200 and 1440 number are about?


    Can you tell the difference? Seriously, Stacey, the issue seems to be the
    size of the ink bubble and
    2, May 1, 2006
    #6
  7. Lee

    Guest

    Lee wrote:
    > Has anyone out there actually compared the printed results of using
    > photoshop bicubic resampling to "interpolate up" the pixels in order to
    > print a larger photo (example: 20x30) than your digital camera would
    > normally be able to support? Were the results better with resampling
    > than just turning it off and sending the photo to the printer with
    > whatever the resolution drops to when you print it at such a large size
    > (example: 8.2megapixel camera 3456x2304pixels. According to Photoshop
    > this image has a resolution of 115 ppi at 20x30. Printers generally
    > recommend 300 dpi minimum). Should I resample this image to
    > 300ppi?Thanks


    I have not done it in Photoshop, but did run a test awhile ago in Paint
    Shop Pro, in a version that allowed choice of bicubic versus linear.
    The bicubic was clearly better for the image I was working with, but
    the best algorithm may well depend on the specific image. By and
    large, however, the bicubic works well.
    , May 1, 2006
    #7
  8. Re: Does bicubic resampling make a noticeable improvement to printedphotos?

    Stacey wrote:
    > bmoag wrote:
    >
    >> Inkjet photo printers print natively at 300dpi.

    >
    > Wrong. Canon native hardware is 600, epson is 720. What do you think those
    > 1200 and 1440 number are about?
    >
    >> Printer manufacturers may cite specs ranging from slightly less to
    >> significantly higher but the fundamental number is 300dpi for the highest
    >> quality ink jet photo printers.

    >
    > What you're reading is people saying 300DPI is what you need to start with
    > before printer driver upsampling to get a good print.
    >
    >> You can set the print DPI smaller: many if not most images printed at 250
    >> dpi are not distinguishable from images printed at 300 dpi.

    >
    > Exactly because the printer driver upsamples it to the native hardware rez
    > before printing.
    >

    Well, you are both wrong, and this illustrates a small number of
    instances where mixing up dpi and ppi makes a difference.

    Sheesh. For all the posts you each make, youdda thunk different.

    --
    John McWilliams
    John McWilliams, May 2, 2006
    #8
  9. Lee

    Stacey Guest

    John McWilliams wrote:

    > Stacey wrote:
    >> bmoag wrote:
    >>
    >>> Inkjet photo printers print natively at 300dpi.

    >>
    >> Wrong. Canon native hardware is 600, epson is 720. What do you think
    >> those 1200 and 1440 number are about?
    >>
    >>> Printer manufacturers may cite specs ranging from slightly less to
    >>> significantly higher but the fundamental number is 300dpi for the
    >>> highest quality ink jet photo printers.

    >>
    >> What you're reading is people saying 300DPI is what you need to start
    >> with before printer driver upsampling to get a good print.
    >>
    >>> You can set the print DPI smaller: many if not most images printed at
    >>> 250 dpi are not distinguishable from images printed at 300 dpi.

    >>
    >> Exactly because the printer driver upsamples it to the native hardware
    >> rez before printing.
    >>

    > Well, you are both wrong, and this illustrates a small number of
    > instances where mixing up dpi and ppi makes a difference.
    >
    >


    So what part of "the printer driver upsamples it to the native hardware rez"
    is wrong? You're argueing semetics here.


    --

    Stacey
    Stacey, May 3, 2006
    #9
  10. Lee

    Stacey Guest

    2 wrote:

    > "Stacey" <> wrote in message
    > news:...
    >> bmoag wrote:
    >>
    >>> Inkjet photo printers print natively at 300dpi.

    >>
    >> Wrong. Canon native hardware is 600, epson is 720. What do you think
    >> those 1200 and 1440 number are about?

    >
    > Can you tell the difference? Seriously, Stacey, the issue seems to be the
    > size of the ink bubble and


    It's not a big difference, but it's there. Qimage does a touch of sharpening
    AFTER upsampling it to the native printer hardware rez which helps a bit.
    If you're starting out with a 150-200DPI file, it's more obvious.

    --

    Stacey
    Stacey, May 3, 2006
    #10
  11. Lee

    2 Guest

    "Stacey" <> wrote

    > [...] What do you think those 1200 and 1440 number are about?


    Aren't those numbers just step motor metrics? Does it really factor directly
    to image resolution? I think not.
    2, May 3, 2006
    #11
  12. Re: Does bicubic resampling make a noticeable improvement to printedphotos?

    Stacey wrote:
    > John McWilliams wrote:
    >>>

    >> Well, you are both wrong, and this illustrates a small number of
    >> instances where mixing up dpi and ppi makes a difference.
    >>

    >
    > So what part of "the printer driver upsamples it to the native hardware rez"
    > is wrong? You're argueing semetics here.
    >

    Didn't opine on driver upsampling.

    But it's simply misleading to folks trying to learn terminologies and
    processes to speak of an image's dpi before it is printed. It's pixels
    per inch until then.

    If this seems as if I'm arguing semantics, you miss the point.

    --
    John McWilliams
    John McWilliams, May 3, 2006
    #12
  13. Lee

    Bill Tuthill Guest

    Lee <> wrote:
    > Has anyone out there actually compared the printed results of using
    > photoshop bicubic resampling to "interpolate up" the pixels in order to
    > print a larger photo (example: 20x30) than your digital camera would
    > normally be able to support? Were the results better with resampling
    > than just turning it off and sending the photo to the printer with
    > whatever the resolution drops to when you print it at such a large size
    > (example: 8.2megapixel camera 3456x2304pixels. According to Photoshop
    > this image has a resolution of 115 ppi at 20x30. Printers generally
    > recommend 300 dpi minimum). Should I resample this image to
    > 300ppi?Thanks


    What printer?

    Some printers, e.g. Epson and HP, are not as good at upsampling
    as "Bicubic Smoother" in Photoshop CS2. (Bicubic Smoother seems
    better than older techniques such as 110% stairstepping.)

    Other printers, such as Lightjet and Durst Lambda, have better
    upsampling algorithms than Photoshop.

    Lanczos can be superior to Bicubic for certain images, and is
    widely available, just not in Photohsop. There are several
    advanced techniques such as Backprojected Jensen-Zhao-Xin Li
    which are better with most images.

    I'm not going to dignify the PPI/DPI argument with a reply
    other than to say that's just unhelpful Photoshop BS.
    The only thing that matters is total pixel size.
    Bill Tuthill, May 3, 2006
    #13
  14. Lee

    Stacey Guest

    John McWilliams wrote:

    > Stacey wrote:
    >> John McWilliams wrote:
    >>>>
    >>> Well, you are both wrong, and this illustrates a small number of
    >>> instances where mixing up dpi and ppi makes a difference.
    >>>

    >>
    >> So what part of "the printer driver upsamples it to the native hardware
    >> rez" is wrong? You're argueing semetics here.
    >>

    > Didn't opine on driver upsampling.


    Well that was the question being discussed. The canon printer is looking for
    a 600ppi file and if not fed one, it creates one. LOTS of people use PPI
    and DPI as being the same and given it's basically a "digital unit per
    inch" whichever term you use, does it really make much difference other
    than semantics?

    >
    > If this seems as if I'm arguing semantics, you miss the point.
    >


    Seems you missed the point of the OP's question!

    --

    Stacey
    Stacey, May 4, 2006
    #14
  15. Lee

    Stacey Guest

    2 wrote:

    > "Stacey" <> wrote
    >
    >> [...] What do you think those 1200 and 1440 number are about?

    >
    > Aren't those numbers just step motor metrics? Does it really factor
    > directly to image resolution? I think not.


    You might have a talk with the guy who writes the Qimage software. He's not
    the only one who suggests to feed these printers these size files. It's
    what the printer driver does if you don't.

    --

    Stacey
    Stacey, May 4, 2006
    #15
  16. Lee

    2 Guest

    "Stacey" <> wrote in message
    news:...
    >2 wrote:
    >
    >> "Stacey" <> wrote
    >>
    >>> [...] What do you think those 1200 and 1440 number are about?

    >>
    >> Aren't those numbers just step motor metrics? Does it really factor
    >> directly to image resolution? I think not.

    >
    > [...] It's what the printer driver does if you don't.


    How do you know that, and regardless, how do you know it actually makes
    700ppi prints?
    2, May 4, 2006
    #16
  17. Lee

    Ed Ruf Guest

    On Thu, 4 May 2006 07:37:37 -0500, in rec.photo.digital "2"
    <> wrote:

    >"Stacey" <> wrote in message
    >news:...
    >>2 wrote:
    >>
    >>> "Stacey" <> wrote
    >>>
    >>>> [...] What do you think those 1200 and 1440 number are about?
    >>>
    >>> Aren't those numbers just step motor metrics? Does it really factor
    >>> directly to image resolution? I think not.

    >>
    >> [...] It's what the printer driver does if you don't.

    >
    >How do you know that, and regardless, how do you know it actually makes
    >700ppi prints?


    Try for yourself.
    http://www.ddisoftware.com/qimage/quality/

    ________________________________________________________
    Ed Ruf Lifetime AMA# 344007 ()
    http://EdwardGRuf.com
    Ed Ruf, May 4, 2006
    #17
  18. Lee

    2 Guest

    "Ed Ruf" <> wrote:

    >>How do you know that, and regardless, how do you know it actually makes
    >>700ppi prints?

    >
    > Try for yourself.
    > http://www.ddisoftware.com/qimage/quality/


    Most excellent! Thanks for that, Ed.
    2, May 4, 2006
    #18
  19. Re: Does bicubic resampling make a noticeable improvement to printedphotos?

    Stacey wrote:
    > John McWilliams wrote:
    >
    >> Stacey wrote:
    >>> John McWilliams wrote:
    >>>> Well, you are both wrong, and this illustrates a small number of
    >>>> instances where mixing up dpi and ppi makes a difference.
    >>>>
    >>> So what part of "the printer driver upsamples it to the native hardware
    >>> rez" is wrong? You're argueing semetics here.
    >>>

    >> Didn't opine on driver upsampling.

    >
    > Well that was the question being discussed. The canon printer is looking for
    > a 600ppi file and if not fed one, it creates one. LOTS of people use PPI
    > and DPI as being the same and given it's basically a "digital unit per
    > inch" whichever term you use, does it really make much difference other
    > than semantics?
    >

    Yes, Stacey, a lot of people confuse PPI and DPI, including you who
    knows better, although now that you're engaged in arguing that point
    with me, I bet such a distinction is lost.

    Pixels of an image and dots of ink sprayed out by a printer are
    different. Way different.

    --
    John McWilliams
    John McWilliams, May 4, 2006
    #19
  20. Lee

    Bill Tuthill Guest

    >>> [...] What do you think those 1200 and 1440 number are about?
    >>
    >> Aren't those numbers just step motor metrics? Does it really factor
    >> directly to image resolution? I think not.

    >
    > You might have a talk with the guy who writes the Qimage software. He's not
    > the only one who suggests to feed these printers these size files. It's
    > what the printer driver does if you don't.


    But http://www.ddisoftware.com/qimage/quality/
    implies that the numbers are 600 (Canon or HP) and 720 (Epson)
    not 1200 and 1440.

    The above webpage reinforces what I posted earlier
    about Photoshop not having the best upsampling algorithms,
    although Bicubic Smoother (as of CS) is better than samples
    on the above webpage done with Photoshop 7.
    Bill Tuthill, May 4, 2006
    #20
    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. Mr.Will

    resampling a digital image to make it bigger?

    Mr.Will, Oct 22, 2003, in forum: Digital Photography
    Replies:
    19
    Views:
    652
    Bill Hilton
    Oct 23, 2003
  2. Christian

    Bayer to RGB bicubic algorithm

    Christian, Jan 29, 2004, in forum: Digital Photography
    Replies:
    5
    Views:
    11,746
    Mike Brodbelt
    Feb 2, 2004
  3. GTO
    Replies:
    16
    Views:
    494
    Mark Johnson
    Oct 11, 2004
  4. Bicubic Interpolation

    , Mar 3, 2006, in forum: Digital Photography
    Replies:
    4
    Views:
    1,371
    Matt Ion
    Mar 4, 2006
  5. John Navas

    Adobe Bicubic Smoother upsize vs Genuine Fractals

    John Navas, Nov 27, 2007, in forum: Digital Photography
    Replies:
    30
    Views:
    2,215
    Bill Tuthill
    Dec 1, 2007
Loading...

Share This Page