re-sampling images before printing

Discussion in 'Digital Darkroom' started by photo_girl|1, Nov 24, 2008.

  1. im slightly confused when it comes to re-sampling images before one goes to print.

    I'm currenty working on my D700 shooting in RAW (native resolution 4256 by 2832), and i print on my Epson 3800
    where i've been told that 360dpi is the best to print at.

    Question is, i find myself mostly printing A3+ size images (13 by 19 inches). When i go to print, do i need to
    change settings in my IMAGE SIZE menu? i tend to just go to page set-up, select the page size.. then i go to
    print with preview select everything and click on scale to fit media...

    am i supposed to do something else in IMAGE SIZE while in photoshop for better results? i thought maybe i should
    open up image size, then in document size - punch in 13by19 inch...... and go from there... perhaps when i do
    that i need to select Bicubic Sharper...

    can someone confirm whether im on the right track....? whether i can just go to print preview and select fit to
    media.. keeping my image at 360dpi... or whether i HAVE to go to IMAGE SIZE, punch in the image size that im
    after and click on bicubic sharper... (been told to always click that whether i upscale or down-scale image)
     
  2. rnt

    rnt

    I've had issues with aliasing when I leave the rescaling to my 3800 (the Scale to Fit option). I now generally rescale to the proper size in Photoshop, print, and then discard the rescaled image (either use the history palette to 'undo' the resize, or don't save the rescaled image if that's all I've done to it). If you rescale within Photoshop you should select the appropriate rescaling algorithm- there's one for upsizing and another for downsizing (among others).
     
  3. You can send whatever you want to the printer and it will resample the image to it's native resolution (720dpi). However, if you rescale in PS, you will get better results. The printer driver is not good at translating vectors such as circles, arcs, and diagonal lines...it will make stepped edges.

    PS will fill the steps and make sharper edges. Use bicubic smoother for upsizing - apply sharpening to the upsized image.

    Use bicubic sharper when down sizing.
     
  4. According to Jeff Schewe, Greg Gorman ,and Mac Holbert at the Epson Print Academy seminar a couple of weeks ago:

    You should be fine by not resampling your images as long as the file you send it has greater than 180ppi resolution.

    So, open the photo in Photoshop (you are working in PsCS3 as I recall). Go Image >image Size and turn off (by unchecking) Resample Photo. this links the Width, Height and Resolution settings in Document Size.

    Now adjust the height or width to the size you want and as long as the Resolution is above 180 pixels/inch you should not see a reduction in image quality. Even the representative from Genuine Fractals parent company acknowledged this. But if f you go below 180 then that is the time to employ Genuine Fractals or other methods of resolution interpolation.

    Now turn the Resample Image tool back on.
     
  5. The resampling filter within the printer is better than what is in ps. Does a great job.
     
  6. ellis, i just tried out what you suggest on an image.... i uncheck the resample image box.. punch in 13 by 19 inch in document size.. and my resolution drops down to 140........... from what i know, i need to be printing at 360 for my epson.. this is not making sense :(
     
  7. jtk

    jtk

    IMO trust the printer..except when making tiny prints (eg 2" sq).
     
  8. >>> from what i know, i need to be printing at 360 for my epson..

    Nah... Anything above 180 or so works great...
     
  9. And the CTO of ImagePrint said >150 in a technical white paper.
     
  10. fair enough, but when i do what ellis suggested i do... my resolution changes to 140.. not even 180 !
     
  11. "from what i know, i need to be printing at 360 for my epson.. this is not making sense :("

    What you know ain't necessarily so

    Why don't you go ahead and make some prints and let us know the results?
     
  12. thank you ellis... should i still go ahead and print even if im getting a resolution UNDER 180 when i re-sample as you had suggested..? you said anything over 180 is fine..... but im getting 140 when i re-sample the way you instructed :)
     
  13. Frankly I don't bother. John is right, trust the printer. I have a complete HP set-up and do only 8x10 working prints for myself, the rest goes off to a prolab. I never had any problems and in fact the 8x10's I do myself are really very good, in colour and even more importantly in b&w as well. So no resampling here.

    Not everything has to be (made) complicated.
     
  14. "but im getting 140 when i re-sample the way you instructed :)"

    If you get 140 ppi and you're printing a 13 x 19, then your image is only 1820 x 2660 pixels (13x140 by 19x140). Did you crop it a lot?
     
  15. no cropping at all...

    \i bourght it in from camera raw... camera's native resolution (my D700) and went straight into what ellis suggested, just to see what the steps are... and thats what happened.... :(
     
  16. i just re-checked, it actually comes up to 149. xxxxx...
     
  17. What does it say in the Image Size window at the top under "Pixel dimensions: Width and Height"?
     
  18. aha, ok i thnk we're on to something alan - my wrong... i've just tried to punch in 13 by 19inches in the pixel size... after i UNCHECK resample image.... when i punch in width, 13 works fine.. but when i go to punch in 19 inches for height.. the width changes to 28.554 inches....

    sorry, didnt realize this was happening... how do i do it so that i punch in exactly what i want?
     
  19. You can't get exactly 13 x 19 inches when you resize a full resolution image because 13 x 19 isn't the same aspect ratio as the sensor which is 2:3. You don't want to put a number into both width and height. If your image is a vertical image ("portrait orientation") and you type in 13 into the width, the height will automatically display it's dimension which should be about 19.5 inches or so. Once you've resized like this you will have to go back and crop to 13 x 19. This is all assuming you have cropped or resised the image earlier.

    You didn't answer about what the pixel dimensions are when your in the Image Size window.
     
  20. great, thank you so much - finally got it and understand it :) phew

    my last question is - a while back i read an article stating that whether i upsize or downsize image to ALWAYS go
    with image sharper... while other schools of thought say that when i upsize to go with bicubic smoother and
    downsize... to go with sharper.. any thoughts on this?
     
  21. ONE more question - what about if i find myself in a situation where the client needs an image on CD to then send out to publications.... im doing a portrait next week for a client .. who needs the image on a CD, to then just send out to random publications that request it (public figure) as i dont know the size etc that the image will be set at.... is there a general setting i should set IMAGE SIZE at....? how should i go about that?
     
  22. I did this test a while back by sending file sizes of different resolutions to the printer. The files that were interpolated to 240 or 360 had no roughness to straight lines......images that were sent at off sizes like 207 and 289 had a rough look to fine lines.

    So, I always interpolate the image to 240dpi or 360dpi prior to sending to the printer, and I've found better results. YMMV.
     
  23. try printing with Qimage. All that resampling...too many numbers to keep in my head.
     
  24. Very interesting. This is just to mark it. Will red when I have time. Karl
     
  25. Here is a handy little tool for determining your printers optimum native resolution by using your eyeballs. Just print it out, making no changes in size or resolution and examine with a magnifying glass or a loupe.

    PrintTest
     
  26. One cause of confusion in this thread seems to be that some people are talking about DPI, and some about PPI.

    PPI is Pixels Per Inch, and refers to the pixel data in the image file.
    DPI is Dots Per Inch, and refers to the physical ink dots used by the printer to convey the image on the paper.

    A Pixel contains full RGB (or CMYK) information in one pixel unit. Dots on the other hand are single color ink dots. To convey RGB
    information, the printer needs to print multiple dots of ink. Hence the PPI figure is a measure of "more expressive" data, and you need
    less PPI for any given DPI figure.

    It seems Epson is recommending 180 PPI resolution for their printers, which sounds like a sensible figure. Assuming this is the
    resolution to aim for with Epson, just check PS is showing you the "Pixels/inch" figure and try to keep that above 180. And ignore the
    DPI figures in the printer.
     
  27. The advice in CS3 help file is to resize the image with *unchecked* resample, and, after that, if the resulting resolution is not satisfactory, to resample the image at the desired resolution
     
  28. I am printing in my 3800, and as Dave, I've got the best results (with straight lines) resampling to 240DPI. I have tried 360DPI but I see no detail difference, so I always use 240DPI.
    Hope it helps.
    Regards, diego.
     
  29. True Diego. I only use 360dpi when the native rez is close to that. If it's closer to 240, I use that. I surprised some people still question this. The results are easy as a sample print away. NO fine art printer would leave the interpolation up to the print driver as it is NOT better than maintaining control yourself in PS.
     
  30. Higher DPI may not give more detail, but sometimes for large prints I use 1200DPI which gives me much better saturation that I want for copies of my paintings of Red Rock Country, Grand Canyon, etc....
     
  31. >>> I surprised some people still question this.

    Because very knowledgeable experts Jeff Schewe, (the late) Bruce Fraser, and others, have run tests and disagree.

    Certainly makes no difference on my 4800...
     
  32. "Because very knowledgeable experts Jeff Schewe, (the late) Bruce Fraser, and others, have run tests and disagree." <br>
    I had a long discussion with Jeff Schewe.... it was clear that he is not able to understand the difference between sampling and dithering.
     
  33. "Because very knowledgeable experts Jeff Schewe, (the late) Bruce Fraser, and others, have run tests and disagree."

    I never believe experts unless I can personally confirm what they say. I'm going to tell you that in this case, they either don't know what to look for or are not very discriminating. I can prove that the printer driver is an inferior method of interpolation to a larger size compared to using PS bicubic smoother, which is inferior to Kneson Imagener, which is inferior to Qimage.

    If you are doing relatively small prints - it will work adequately - but not as well as PS bicubic interpolation. I make 18x29-inch prints from a 10 megapixel camera. In my testing, I'm sending 137ppi information to the printer driver - it does NOT work well. Sending the same data to Qimage makes a visibly better print. When evaluated with a 5x loupe there in NO contest - the printer driver sucks... in comparison. I can show anyone the difference.

    The printer driver WILL GIVE STEPPED EDGES on circles, arcs, diagonal lines. It CANNOT REPRODUCE VECTORS cleanly. Photoshop will fill in the steps, but in the process, makes a gradient edge of between 2-4 pixels meaning the edge rolls off and is not a sharp line. Kneson Imagener will make sharp edges but can get confused on large areas of solid color intersected by crossing lines (think of power lines against a blue sky). In that case, the software will pixellate areas very slightly as it seems to want to create a texture between the lines. Genuine Fractals will make small details into solid objects. I rejected GF V after making only two prints with it and would not recommend it to anyone.

    Qimage, will make sharp edges, no steps, no roll off and preserve details without pixellation. The Pyramid interpolation routine is absolutely and with NO question far, far superior to any other interpolation method tested.

    I have no idea what Jeff Schewe is looking at, but anyone with a 5x loupe and a bit of discrimination can easily see the stepped edges with the printer driver. If you can't - you don't qualify as an "expert."
     
  34. "I had a long discussion with Jeff Schewe.... it was clear that he is not able to understand the difference between sampling and dithering."

    If that's a true statement - no wonder he thinks the printer driver works well for interpolation.
     
  35. I too get a bit confused RE dpi and ppi :) On Monday I made some 20x30 inch prints from my 400D (file size is 3888x2592 for large JPEG). I didn't resample or uprez the images. The print output was set to @500dpi according to the guy who printed them and the output was great!
     
  36. i thought that once one unchecks resample image, types in the required document size... then checks the resample image box again, then one needs to go straight to OK. didnt realize that i can then type in a preferred ppi... either 240 or 360. I realize that i end up with a massive file in this case, but as I'm doing this only for print and not Saving that image, then i can just delete it, keeping my Master file. Correct?
     
  37. I've found that 288 ppi file resolution also works well with most of the Epson printers I've used, when used with the Epson driver but not necessarily with other drivers - for example OPM has a sweet spot at 300 ppi on a 2200 but not at 288 or 360. That is based on the degree of pixel-to-pixel similarity between the file and the printed image when printing at '1440 dpi' and '2880 dpi' settings.

    Best,
    Helen
     
  38. "I too get a bit confused RE dpi and ppi :) On Monday I made some 20x30 inch prints from my 400D (file size is 3888x2592 for large JPEG). I didn't resample or uprez the images. The print output was set to @500dpi according to the guy who printed them and the output was great!"
    <br><br>
    You printed sending 129.6 PPI .
    <br>
    129.6=3888/30=2592/20
    <br><br>
    The guy is wrong. PPI is the correct term.
     
  39. >I had a long discussion with Jeff Schewe.... it was clear that he is not able to understand the difference between
    sampling and dithering.

    Care to remind me of what life you and I had a discussion (or what forum and what screen name)...

    Also care to refine your statement regarding what you are talking about? I pretty much know how the previous Epson
    printers create their droplets using an error diffusion method. The new 7900/9900 printers use an all new method that is
    considerably more exotic. But the EPson driver (prior to the 79/9900) doesn't use a "resampling" interpolation to create
    the screening...if anything, it's like a sieve with the fineness of the sieve depending on the print quality settings in the
    driver. Understand that inkjet printers don;t print with "dots" they print with "droplets" in the picoliter range–a
    measurement of volume not size. The pro line that Epson currently sells goes down to 3.5 picoliters and THAT does not
    translate to DPI or PPI.

    The "effective" resolution of Epson printers is said to be 360DPI (the new 79/9900 actually have 360 nozzles per inch,
    the older printers had 180). The HP and Canon printers are said to be 300DPI but that simply doesn't translate into REAL
    RESOLUTION of the droplet size at all. The whole thing is either far more or a lot less complicated as people make out.

    Then factor in the effective resolving power of human vision is about 1.2 arcminute per line pair, or a 0.35 mm line pair,
    at 1 m according to wikipedia, but that doesn't really tell the whole story either since photos are not high contrast line
    pairs. The reason I had said that is you have between 180-480PPI, don't resample just resize is that the actual ability to discern image detail is a function of the viewing distance (which goes along with the way human vision works) so, the
    closer you hold a print to your eyes, the more resolution you need to fool the eye into seeing "continuous tone" and not
    dots (or droplets).

    Kodak, when they worked on their dyesub printers determined that for the print size of their printers they needed 204
    DPI. I always though that was a bit low. But, if you have a reasonable digital capture (6+MP) then a minimum of 180 will
    result in a bigger print and the max of 480 will produce a smaller print–without having to "resample". But, if you DO need
    to resample digital captures, going 200% isn't hard and a really good capture can prolly go up 400%.

    So far, I've seen a lot of confusion (and myth as well as misquotes) in this thread. The bottom line is, do what you are
    doing, make a test print and see f you can see any differences. If you can't fine and dandy, if you can see differences,
    then learn how to do something about it. The last thing I'll mention is all the "resolution" discussion also comes down to
    how you prepare the image for printing and output sharpening. But, that's another whole discussion :~)
     
  40. "Care to remind me of what life you and I had a discussion (or what forum and what screen name)... "
    <br><br>
    http://luminous-landscape.com/forum/index.php?showtopic=23167&pid=174711&mode=threaded&start=#entry174711
    <br><br>
    and
    <br><br>
    http://luminous-landscape.com/forum/index.php?showtopic=23125&pid=176012&mode=threaded&start=#entry176012
    <br><br>
    This is my last post:<br>
    "If the sampling is at 360 dpi (PPI is the correct term) and the image is not 360 PPI, this is a resampling.
    <br>
    Dithering as I tried to explain is the process to simulate the color of the sampled pixel.
    <br>
    You have a strange idea about dithering."
     
  41. You were quoting an Epson PDF which was essentially wrong. The driver doesn't "sample", it takes whatever data it's
    given, assumes an effective 360DPI (720 if Finest Detail is checked–which not suggested for photos) and then uses an
    error diffusion to dither the image data and derive a droplet map.
     
  42. "You were quoting an Epson PDF which was essentially wrong. The driver doesn't "sample", it takes whatever data it's given, assumes an effective 360DPI (720 if Finest Detail is checked–which not suggested for photos)"
    <br><br>
    Yes,I see "every thing that doesn't match what you think is wrong."
    <br><br>

    There is a strong conflict between "it takes whatever data it's given'....... and "assumes an effective 360 DPI".
    <br>
    I hope you mean PPI.
    <br>
    I gave evidence of the driver PPI setting when you change the printer quality and invited you to check the driver SDK from Epson.
    <br>
    How do you explain that for low PPI values you can see the pixels on print?
    <br><br>
    You prefer to think you are right, you are free to do that.
    <br>
    But your "truth" is not a "truth".
    <br>
    Resampling is "science" not "art", as you think.
     
  43. Jeff,

    I've done the sample printing myself on an Epson 3800, 4800, and 7600. Feeding the printer resolutions like 129, 167, 273, and then feeding it 240 or 360 definitely looks different. The 240 or 360 provide smoother lines with no jagged edges.

    Feel free to quote whatever paper you like, but the fact remains that better results can be obtained by feeding the printer 180, 240, or 360 dpi output. I've done this test countless times with people who argue incessantly about it....and always show the same thing......240 or 360 is better.

    Now in all fairness, this is if you're looking at sharp, perfect lines, from about 8" from glossy paper. Will it show up in real life normal photography, I doubt it. Very few images I've taken show a problem.....but why get lazy.....simply upsample yourself and avoid the problem altogether.

    That said, I'm lazy once in while myself!

    Regards,
     
  44. Does anybody else find that 288 ppi also works well?

    Thanks, Helen
     
  45. Helen,
    if you have Windows or can run a Windows application, go to PhotoResampling site and download PrinterData.<br>
    It is free. <br>

    Changing the printer prefereces you can get the PPI value.

    If you want to know how good is 288 PPI you have to compute the scale factor Sf= printerPPI/288. <br>

    At this point, take a bunch of destination pixel positions xd, for example 0,1,2,3,4,5,6 for each one compute the source pixel position xo as:<br>

    xo=(xd/Sf) <br>
    take the integer part of the division .<br><br>
    In this way you can know the mapping between source and destination pixels.

    For example, if the printer PPI is 360, you get: <br>
    0,0,1,2,3,4,4 <br>

    for the above 7 destination pixels.
    <br>
    It means that a pixel is used 2 times, then 3 pixels are used.......and so on.

    Is it good? Difficult to say.
    There are 2 problems:<br>
    - some pixels are picked more than 1 time<br>
    - the replicated pixels are not symmetric<br>

    It is visible on print? hard to say, it depends on the image. <br>

    Generally the better think to do is to use the printer PPI value.<br>
    Often the PPI value/2 is good enough (every pixel is used 2 times).
    Different values may produce good print but it is not guaranted.
     
  46. Yes Helen, kind of. When we still had an Epson inkjet
    (boy are we happy to be rid of that albatross!)
    144 dpi worked well, which as those of us on photo.net who own a calculator can tell,
    is half of 288. Any divisor of print-resolution works well.

    Some high-end printers such as the Lightjet might have better resampling algorithms
    than Photoshop, but Epson does not.
     
  47. Thanks for the replies.

    Jacopo - I don't need to know how good 288 ppi is because I've already done extensive testing. It suggests that at 288 ppi I can get a nearly perfect file-pixel to print-pixel match - in many cases better than that which occurs when a 360 ppi file is sent to the printer. 288 ppi allows a 5x5 matrix at 1440 dpi, though that is not exactly what happens in both the horizontal and vertical (it has also changed with printer generations, of course). As I mentioned earlier this only happens with the Epson driver, third-party RIPs do not produce the same result.

    Best,
    Helen
     
  48. Helen, do you theorize that the 5x5 matrix works best
    because of the 5 non-black color droplets (CcMmY) that can be combined?
     
  49. I do recall reading an article on sharpening by Jeff Schewe where he states that the amount of output sharpening is dependent upon the final media used AND the final image resolution, ie. 240 PPI compared to 350 PPI.

    For complete info read Jeff's article here:
    http://www.pixelgenius.com/tips/schewe-sharpening.pdf

    Don Bryant
     
  50. This site...

    http://www.silverace.com/dottyspotty/issue9.html

    and a sample page below linked at the bottom of the same page shows and explains what differences in picoliter amounts look
    like and how an inkjet's dithering pattern can hide quite a bit of stairstepped edges lowering the resolution. Experimentation
    with your own printer will have to be conducted on what you can get away with. On an old Epson EX I had I determined I could
    start seeing stairstepped pixelation on image resolution starting at 150 ppi and below.

    http://www.silverace.com/dottyspotty/issue93.jpg
     
  51. Helen,

    I just tried the 288 on a test image with ultra fine lines and it does indeed work well. It appears I can now use 240, 288 and 360 without worry.

    Been a fun thread. I'm still surprised some people argue the issue when all they have to do is make a print!
     
  52. Bill,

    Rather in the same vein that Dave is following, I'm wary about getting too theoretical - I'm just musing about what the reason could be. It seems wrong to base things solely on 360 blocks or whatever per inch as the basic unit into which everything else has to fit. 1440/4, 1440/5 and 1440/6 all seem to work nicely.

    Best,
    Helen
     

Share This Page