question on dpi scalling in Photoshop

Discussion in 'Digital Photography' started by Apkesh, Oct 25, 2004.

  1. Apkesh

    Apkesh Guest

    Hi,
    I have a photo which I took it at 2048X 1360 pixels with a Canon 10D.
    I love the photo and want to enlarge it by sending it to an online
    lab. When I opened it in Photoshop, it opened at roughly 11 inches by
    7.5 inches. If I wish to *enlarge* this to 10X15, I guess I'll first
    need to increase it's DPI?

    I made the image size 10X15 and then changed the DPI to 320 (no
    particular reason how I got to that number, but just wanted to
    increase it). The image looks good on the screen, but I am not sure
    how it will turn out in print.

    Does anyone know how and if DPI settings make a difference? Thanks!

    apkesh
     
    Apkesh, Oct 25, 2004
    #1
    1. Advertising

  2. "Apkesh" <> wrote in message
    news:...
    > Hi,
    > I have a photo which I took it at 2048X 1360 pixels with a Canon 10D.
    > I love the photo and want to enlarge it by sending it to an online
    > lab. When I opened it in Photoshop, it opened at roughly 11 inches by
    > 7.5 inches. If I wish to *enlarge* this to 10X15, I guess I'll first
    > need to increase it's DPI?


    You can DECREASE the number of dots per inch, keeping the same number of
    dots. In particular, 2048/15=136 dots per inch. To do this, make sure
    "resample" is not checked when you go to Image Size.

    Or you can RESAMPLE the image (increase the total number of dots) to keep
    the same dpi. To do this, make sure "resample" is checked.
     
    Michael A. Covington, Oct 25, 2004
    #2
    1. Advertising

  3. Apkesh

    Bob Williams Guest

    Apkesh wrote:
    > Hi,
    > I have a photo which I took it at 2048X 1360 pixels with a Canon 10D.
    > I love the photo and want to enlarge it by sending it to an online
    > lab. When I opened it in Photoshop, it opened at roughly 11 inches by
    > 7.5 inches. If I wish to *enlarge* this to 10X15, I guess I'll first
    > need to increase it's DPI?
    >
    > I made the image size 10X15 and then changed the DPI to 320 (no
    > particular reason how I got to that number, but just wanted to
    > increase it). The image looks good on the screen, but I am not sure
    > how it will turn out in print.
    >
    > Does anyone know how and if DPI settings make a difference? Thanks!
    >
    > apkesh


    When PS opens a 2048x1360 pixel image, it usually displays it at 72
    pixels/inch and the rulers will show it to be 28.4" x 18.9"
    If you want to create a 15x20 print @ 320 pixels/inch (not DPI), just
    double click on the crop tool and set the width and height to 15x20 and
    set the resolution to 320 ppi. When you crop your picture the way you
    like it, the image will be EXACTLY 15x20 inches @ 320 pixels/inch.
    PS does all the necessary resampling automatically.
    Save the image in the highest quality .jpeg (any processor can handle
    jpegs) and send it off.
    Bob Williams
     
    Bob Williams, Oct 25, 2004
    #3
  4. Apkesh

    Apkesh Guest

    okay, so let me get this straight- Even though I've taken the picture
    at 2045 X 1360 pixels (180 DPI), I can still achieve the *same* or
    *better* quality by increasing the DPI count and enlarging the image?
    Is increasing the dpi to 320 any better than setting it to 136
    (2045/15=136). And if so, is there a rule of thumb to stick to in
    increasing the number (Obviously, I would assume changing the dpi to
    800 and keeping the size the same will not make any difference, or
    will it?)

    apkesh
     
    Apkesh, Oct 25, 2004
    #4
  5. "Apkesh" <> wrote in message
    news:...
    > okay, so let me get this straight- Even though I've taken the picture
    > at 2045 X 1360 pixels (180 DPI), I can still achieve the *same* or
    > *better* quality by increasing the DPI count and enlarging the image?
    > Is increasing the dpi to 320 any better than setting it to 136
    > (2045/15=136).


    Yes, very slightly, because your computer will interpolate (fill in)
    additional pixels and make everything look smoother, but the file will be
    more than 4 times as large (320/136 squared).

    > And if so, is there a rule of thumb to stick to in
    > increasing the number (Obviously, I would assume changing the dpi to
    > 800 and keeping the size the same will not make any difference, or
    > will it?)


    No print needs appreciably more than 300 dpi. On the other hand, 150 dpi is
    sufficient, and I think 136 is sufficient for a relatively large print.

    Also, note that unlike a photographic enlargement, digital enlargement never
    brings out detail that was not in the original picture.

    On the other hand, if you resample to reduce the total number of pixels in
    the picture (e.g., if you told it to reduce the dpi while keeping the
    picture the same size), you are throwing away detail which you can never get
    back.

    So...

    Change dpi without resampling = change size of picture without changing
    number of pixels in it. This is safe and easy to undo, because the actual
    image hasn't been altered.

    Resampling changes number of pixels in picture. Resampling alters the
    image. It can enlarge an image more smoothly than merely changing the dpi,
    but it can also throw away detail if you reduce the image.

    Always keep the original file separately if you want to experiment with
    resampling.
     
    Michael A. Covington, Oct 25, 2004
    #5
  6. Apkesh

    Bob Williams Guest

    Apkesh wrote:
    > okay, so let me get this straight- Even though I've taken the picture
    > at 2045 X 1360 pixels (180 DPI), I can still achieve the *same* or
    > *better* quality by increasing the DPI count and enlarging the image?
    > Is increasing the dpi to 320 any better than setting it to 136
    > (2045/15=136). And if so, is there a rule of thumb to stick to in
    > increasing the number (Obviously, I would assume changing the dpi to
    > 800 and keeping the size the same will not make any difference, or
    > will it?)
    >
    > apkesh


    Resampling upward to 320 ppi does not add any new INFORMATION to the
    image but tends to reduce pixelation by filling in interpolated colors
    between pixels resulting in a somewhat smoother image. Personally, I
    would not resample to 320, but that is what you said you wanted to do. I
    would, however resample to about 240 ppi because I've not seen much
    improvement in image quality by going higher than that.
    PS handles bicubic resampling up to 2X very nicely.
    That is all that you are asking it to do in going from 136 to 240 ppi.
    I'm not sure where the point of diminishing returns is. I've never
    resampled to greater than 2X.
    Bob Williams
     
    Bob Williams, Oct 25, 2004
    #6
  7. (Apkesh) writes:
    >okay, so let me get this straight- Even though I've taken the picture
    >at 2045 X 1360 pixels (180 DPI), I can still achieve the *same* or
    >*better* quality by increasing the DPI count and enlarging the image?


    It all depends on your printer driver, if you're printing. The printer
    driver will have to resample the image from whatever DPI setting you
    provide the image in to the actual DPI setting that the printer prints
    in. If it does at least as good a job at this as Photoshop, then doing
    part of the scaling in Photoshop and leaving the rest for the printer
    driver is pointless; it just takes extra time and inflates the amount of
    data sent to the printer driver.

    On the other hand, it might be worthwhile if you use Photoshop to do a
    better job of resizing than the printer driver itself (e.g. by enlarging
    and then doing a carefully-selected amount of unsharp mask).

    Just try printing an image using both methods. If you can't see any
    difference, use the fastest and simplest method.

    Dave
     
    Dave Martindale, Oct 25, 2004
    #7
  8. Apkesh

    Bill Hilton Guest

    >From: (Apkesh)

    >okay, so let me get this straight- Even though I've taken the picture
    >at 2045 X 1360 pixels (180 DPI), I can still achieve the *same* or
    >*better* quality by increasing the DPI count and enlarging the image?


    No, you cannot get the same or "better" quality by resizing.

    >Is increasing the dpi to 320 any better than setting it to 136
    >(2045/15=136).


    What's happening with the printer is this ... it wants a certain ppi value
    internally, say 360 ppi, and if your input file is not that size then the
    printer software will interpolate to get there. Unfortunately most consumer
    inkjets use a mediocre resizing mode, 'nearest neighbor'. If your input file
    is fairly close to 360 ppi then you'll get good results, if it's no where close
    you won't get a very good looking print. I think 240, maybe 200, maybe even
    180 ppi will get resized fairly well with most printers (that's apparently why
    Canon chose 180 ppi, it's probably the lowest ppi # that still gives decent
    prints).

    If your file is TOO small you are better off interpolating yourself, using
    something like Stair Interpolation (110% steps) or if you have CS use 'bicubic
    smoother' or try something like Genuine Fractals if you have a free copy.
    Basically most people with reasonable Photoshop skills can do a better job of
    rezzing up a file and sharpening it with USM than the printer driver does with
    files below a certain ppi threshold.

    >And if so, is there a rule of thumb to stick to in
    >increasing the number


    Run a few tests and see ... for grins reduce a copy of your file to 90 ppi at
    the print size and print that, print it at the native (un-resampled) size of
    180 ppi and also resample up to 240 or 300 ppi and also 360 ppi and print all
    four and compare. You should see that 90 ppi prints poorly, you might be
    surprised at how well 180 ppi prints compared to 360 though.

    >(Obviously, I would assume changing the dpi to
    >800 and keeping the size the same will not make any difference, or
    >will it?)


    Bad idea. The highest native printer ppi that I'm aware of is 400 for the
    Durst Lambda, most do fine with 300-360 ppi (or even 240). There is no reason
    to resample to 800 ppi ...

    Bill
     
    Bill Hilton, Oct 25, 2004
    #8
  9. Apkesh

    Guest

    Bill Hilton <> wrote:
    >>From: (Apkesh)


    >>okay, so let me get this straight- Even though I've taken the picture
    >>at 2045 X 1360 pixels (180 DPI), I can still achieve the *same* or
    >>*better* quality by increasing the DPI count and enlarging the image?



    > If your file is TOO small you are better off interpolating yourself,
    > using something like Stair Interpolation (110% steps) or if you have
    > CS use 'bicubic smoother' or try something like Genuine Fractals if
    > you have a free copy.


    Or Qimage. http://www.ddisoftware.com/qimage/

    > Bad idea. The highest native printer ppi that I'm aware of is 400
    > for the Durst Lambda, most do fine with 300-360 ppi (or even 240).


    You're now aware of the Epson printers -- they run at 720 ppi. I have
    no doubt about this, as I've measured it.

    Andrew.
     
    , Nov 9, 2004
    #9
  10. Apkesh

    Owamanga Guest

    On Tue, 09 Nov 2004 18:07:00 -0000, lid
    wrote:

    >Bill Hilton <> wrote:
    >>>From: (Apkesh)

    >
    >>>okay, so let me get this straight- Even though I've taken the picture
    >>>at 2045 X 1360 pixels (180 DPI), I can still achieve the *same* or
    >>>*better* quality by increasing the DPI count and enlarging the image?

    >
    >
    >> If your file is TOO small you are better off interpolating yourself,
    >> using something like Stair Interpolation (110% steps) or if you have
    >> CS use 'bicubic smoother' or try something like Genuine Fractals if
    >> you have a free copy.

    >
    >Or Qimage. http://www.ddisoftware.com/qimage/
    >
    >> Bad idea. The highest native printer ppi that I'm aware of is 400
    >> for the Durst Lambda, most do fine with 300-360 ppi (or even 240).

    >
    >You're now aware of the Epson printers -- they run at 720 ppi. I have
    >no doubt about this, as I've measured it.


    Please elaborate on your Epson claim. Which model, and how exactly did
    you measure it?

    --
    Owamanga!
     
    Owamanga, Nov 9, 2004
    #10
  11. Apkesh

    Guest

    Owamanga <> wrote:
    > On Tue, 09 Nov 2004 18:07:00 -0000, lid
    > wrote:


    >>Bill Hilton <> wrote:
    >>>>From: (Apkesh)

    >> The highest native printer ppi that I'm aware of is 400
    >>> for the Durst Lambda, most do fine with 300-360 ppi (or even 240).

    >>
    >>You're now aware of the Epson printers -- they run at 720 ppi. I have
    >>no doubt about this, as I've measured it.


    > Please elaborate on your Epson claim. Which model, and how exactly did
    > you measure it?


    It's pretty easy to do. Go to
    http://www.ddisoftware.com/qimage/quality/ and download the test
    image. Print it. Inspect the 720 ppi grating -- you'll need a
    magnifying glass.

    Andrew.
     
    , Nov 10, 2004
    #11
  12. Apkesh

    Owamanga Guest

    On Wed, 10 Nov 2004 11:47:08 -0000, lid
    wrote:

    >Owamanga <> wrote:
    >> On Tue, 09 Nov 2004 18:07:00 -0000, lid
    >> wrote:

    >
    >>>Bill Hilton <> wrote:
    >>>>>From: (Apkesh)
    >>> The highest native printer ppi that I'm aware of is 400
    >>>> for the Durst Lambda, most do fine with 300-360 ppi (or even 240).
    >>>
    >>>You're now aware of the Epson printers -- they run at 720 ppi. I have
    >>>no doubt about this, as I've measured it.

    >
    >> Please elaborate on your Epson claim. Which model, and how exactly did
    >> you measure it?

    >
    >It's pretty easy to do. Go to
    >http://www.ddisoftware.com/qimage/quality/ and download the test
    >image. Print it. Inspect the 720 ppi grating -- you'll need a
    >magnifying glass.
    >
    >Andrew.


    Thanks for the link, and as I suspected, the first test is irrelevant.

    ....unless the world want to print their images in two-tone.

    Most of us are interested in full color 16-24 bit images, and so the
    pixel has to be enlarged to give the printer room to dither. DOTS per
    inch does not equal PIXELS per inch. Therefore, the effective PPI of
    your epson is not the 720ppi you see in BW with this test.

    In other words, this first test allows the pixel to be only pure white
    or pure black. That's a 1 bit color resolution. To represent a pixel
    that is 20.5% cyan 18.0% yellow and 76.5% magenta the printer needs to
    be able to either change the droplet size (and epson printers
    currently support only 6 different droplet sizes) or use a group of
    dots to form a dithered pixel.

    Using 6 ink tanks and 7 droplet sizes (6 + no droplet) it can approach
    a 5 1/2 bit color depth for a single pixel. But we want 24 bit color,
    so we are going to have to use at least 5 of those dots to form a
    single pixel. Suddenly 720ppi becomes 144dpi.

    (The above math isn't perfect because the color gamut of two sets of
    those tanks overlap each other, as does the black overlap a mix of 3
    of the other tanks somewhat. In reality, you need more than 5 dots per
    pixel)

    I don't have my Epson 1270 with me right now to see the test result of
    the second test (the bug photo). I presume you see a significant (or
    possibly subtle) difference between the 720ppi image and the 360ppi...
    I am sure this depends on the color difference between a pixel and
    it's surrounding pixel. Ie will depend on the image.

    I will however switch to downsizing images for printing to 720dpi at
    the designated print size prior to sharpening rather than 300dpi
    because we should be able to notice the difference in monochromatic
    areas of the image. Such a change costs nothing, and is the resolution
    that the printer driver works at, so it only makes sense.

    --
    Owamanga!
     
    Owamanga, Nov 10, 2004
    #12
  13. Apkesh

    bob Guest

    Owamanga <> wrote in
    news::

    > Using 6 ink tanks and 7 droplet sizes (6 + no droplet) it can approach
    > a 5 1/2 bit color depth for a single pixel. But we want 24 bit color,
    > so we are going to have to use at least 5 of those dots to form a
    > single pixel. Suddenly 720ppi becomes 144dpi.
    >


    I understand the concept in general (you got the units backwards though),
    but the pixels aren't linear. Since you don't need a 5x5 matrix of dots it
    shouldn't be so bad. Square root of 5 is about 2.25, which ends up at about
    300ppi

    Bob

    --
    Delete the inverse SPAM to reply
     
    bob, Nov 10, 2004
    #13
  14. Apkesh

    Guest

    Owamanga <> wrote:
    > On Wed, 10 Nov 2004 11:47:08 -0000, lid
    > wrote:


    >>Owamanga <> wrote:
    >>> On Tue, 09 Nov 2004 18:07:00 -0000, lid
    >>> wrote:

    >>
    >>>>Bill Hilton <> wrote:
    >>>>>>From: (Apkesh)
    >>>> The highest native printer ppi that I'm aware of is 400
    >>>>> for the Durst Lambda, most do fine with 300-360 ppi (or even 240).
    >>>>
    >>>>You're now aware of the Epson printers -- they run at 720 ppi. I have
    >>>>no doubt about this, as I've measured it.

    >>
    >>> Please elaborate on your Epson claim. Which model, and how exactly did
    >>> you measure it?

    >>
    >>It's pretty easy to do. Go to
    >>http://www.ddisoftware.com/qimage/quality/ and download the test
    >>image. Print it. Inspect the 720 ppi grating -- you'll need a
    >>magnifying glass.


    > Thanks for the link, and as I suspected, the first test is irrelevant.


    > ...unless the world want to print their images in two-tone.


    > Most of us are interested in full color 16-24 bit images, and so the
    > pixel has to be enlarged to give the printer room to dither. DOTS per
    > inch does not equal PIXELS per inch.


    You seem to be confused between DPI and PPI. To achieve 720ppi, the
    Epson printers have to put down dots at a much higher resolution.

    > Therefore, the effective PPI of your epson is not the 720ppi you see
    > in BW with this test.


    Sure it is.

    The distinction here is between the resolution at which data is sent
    to the printer driver and the dither pattern the printer driver
    generates internally. In order to achieve 720ppi, The dither pattern
    is generated at 2880dpi.

    > In other words, this first test allows the pixel to be only pure white
    > or pure black. That's a 1 bit color resolution.


    No, that's wrong, because your assumptions are wrong. The dither
    pattern is much finer than 720ppi. You are confusing DPI with PPI.

    > I don't have my Epson 1270 with me right now to see the test result
    > of the second test (the bug photo). I presume you see a significant
    > (or possibly subtle) difference between the 720ppi image and the
    > 360ppi... I am sure this depends on the color difference between a
    > pixel and it's surrounding pixel. Ie will depend on the image.


    I did a comparison between a 7600 and a 1270. It's at
    http://www.littlepinkcloud.com/restest-2.jpg. As you may guess, the
    1270 is the one on the right. The 7600, as you might expect, is much
    more precise, although it might be that the dye ink is spreading more
    than the pigment.

    IOW, although I'm quite happy to believe the 1270 really does have a
    native resolution of 720ppi, you my not see the difference.

    Andrew.
     
    , Nov 11, 2004
    #14
    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. Dryb66

    Question - Boosting DPI

    Dryb66, Apr 8, 2004, in forum: Digital Photography
    Replies:
    9
    Views:
    494
    Paul J Gans
    Apr 11, 2004
  2. DS

    Should File DPI Match Printer DPI?

    DS, Jul 5, 2004, in forum: Digital Photography
    Replies:
    8
    Views:
    646
  3. John

    Question regarding DPI in digital cameras

    John, Oct 13, 2004, in forum: Digital Photography
    Replies:
    8
    Views:
    298
    Linda_N
    Oct 14, 2004
  4. Boooger

    96 dpi x 96 dpi?

    Boooger, Nov 29, 2005, in forum: Digital Photography
    Replies:
    8
    Views:
    2,616
    Bill Hilton
    Nov 29, 2005
  5. Replies:
    7
    Views:
    2,025
    Paul J Gans
    Apr 30, 2006
Loading...

Share This Page