Luminous Landscape Forum
Raw & Post Processing, Printing => Printing: Printers, Papers and Inks => Topic started by: David Watson on October 06, 2011, 10:46:27 am
-
I have read lost of stuff about the resolution required for printing. Some of it seems contradictory and I would appreciate some advice.
I use Epson 3880 and 7900 printers printing through Imageprint 8. I have been told that I will get best results if I set the image resolution to 360dpi but I am unable to tell the difference (other than print size) between an image with the resolution set to 300dpi and one set to 360dpi in PS CS5. I am aware that the RIP may have its own ideas about what DPI it should be printing at but I would have thought that there would have been a noticeable difference.
Furthermore a well respected fine art printer in London (printing on Epson printers) told me he sets the image resolution to 600 DPI.
Is there a definitively correct answer to this?
-
Hi David,
Read this thread: http://www.luminous-landscape.com/forum/index.php?topic=54798.0.
Then if you want more, try searching for "resolution" in the search box. There are many past threads on this subject.
Neil
-
Thank you
Question answered
-
I get the best results using 353.5 ppi. It's divisible by something.
-
Read this thread: http://www.luminous-landscape.com/forum/index.php?topic=54798.0.
Then if you want more, try searching for "resolution" in the search box. There are many past threads on this subject.
Based on that thread (and a natural curiosity) I decided to do some tests and write an article for DPP magazine...the net/net result is that if your native image resolution (at the print dimensions) puts the PPI below 360 (for Epson, 300 for HP & Canon) upsample to 360 PPI (300 PPI) before printing and then do your output sharpening (easy in LR).
If the native rez is above 360 but below 720 PPI (600 PPI for HP & Canon) upsample to 720 PPI (or 600 PPI) and then output sharpen.
The advantages are visible to the naked eye (if you know what to look for) and generally involve a high contrast diagonal or circle and/or super high frequency texture.
Test it yourself (I did before writing the article).
It's easy to upsample and then output sharpen in Lightroom...more difficult (but doable) in Photoshop...
-
If the native rez is above 360 but below 720 PPI (600 PPI for HP & Canon) upsample to 720 PPI (or 600 PPI) and then output sharpen.
Jeff, do you recommend that approach even if the native resolution is only just above 360? I seem to find I have images which work out at about 420ppi, which seems a rather long way below 720.
Jeremy
-
Jeff, do you recommend that approach even if the native resolution is only just above 360? I seem to find I have images which work out at about 420ppi, which seems a rather long way below 720.
If you are printing out to an Epson pro printer whose driver has a Finest Detail setting (the 3880 and above including the 78/9800, 7890/9890, 79/9900 and 4900), yes...the main reason is the ability to upsample and then sharpen and print using Finest Detail (which reports itself as a 720 PPI device)...for the consumer printer it's not an open and shut case as it it for the pro printers....
-
If you are printing out to an Epson pro printer whose driver has a Finest Detail setting (the 3880 and above including the 78/9800, 7890/9890, 79/9900 and 4900), yes...the main reason is the ability to upsample and then sharpen and print using Finest Detail (which reports itself as a 720 PPI device)...for the consumer printer it's not an open and shut case as it it for the pro printers....
Thanks - how about a 3800?
Jeremy
-
Thanks - how about a 3800?
Jeremy
That's my question, too.
My 3800 does have a "finest detail" setting under "quality options" in the print settings. When I get a chance, I intend to try printing one at 360 and another at 720 to see if I can see the difference.
Of course, if Jeff answers first, I'll take his word for it.
Eric
-
Thanks - how about a 3800?
As long as you have a Finest Detail button as an option, yes...the R3000 doesn't so without more testing I'm not sure you'll get much advantage to upsampling to 720 since the print pipeline will end up downsampling it back down to 360 PPI.
In the case of Canon and HP, the highest 600 PPI should report the printer as a 600 PPI device.
-
As long as you have a Finest Detail button as an option, yes...the R3000 doesn't so without more testing I'm not sure you'll get much advantage to upsampling to 720 since the print pipeline will end up downsampling it back down to 360 PPI.
Thanks again. I'll have a look.
Jeremy
-
Thanks for that, Jeff. Very helpful.
Eric
-
So you would never downsample unless the image's original resolution was above 720 dpi?
If my image started at 420 dpi, I've been downsampling to 360 dpi and sending it off to the Epson 3800 to print at 360 dpi native. My thinking was that it's better to lose 60 dpi of IQ from a PS-driven downsample than it is to lose 300 dpi of IQ from the printer driver's upsample interpolation, regardless of post-sampling sharpening.
If I understand you, it's better to upsample from 420 dpi to 720 dpi and send off to the printer preset to finest detail (720dpi)because it makes print sharpening easier/better? And this higher sharpening IQ more than offsets the alternative downsampling degredation?
-
So you would never downsample unless the image's original resolution was above 720 dpi?
If my image started at 420 dpi, I've been downsampling to 360 dpi and sending it off to the Epson 3800 to print at 360 dpi native. My thinking was that it's better to lose 60 dpi of IQ from a PS-driven downsample than it is to lose 300 dpi of IQ from the printer driver's upsample interpolation, regardless of post-sampling sharpening.
If I understand you, it's better to upsample from 420 dpi to 720 dpi and send off to the printer preset to finest detail (720dpi)because it makes print sharpening easier/better? And this higher sharpening IQ more than offsets the alternative downsampling degredation?
That PS driven downsampling may not be so good as you expect it to be. If it was done by Qimage with its adjustable anti-aliasing filter + smart print sharpening then that step may be wise depending on the actual print quality difference possible on the paper used. If there is a substantial difference visible in the print between the printer quality settings that ask for either 360 or 720 PPI rendering resolution (based on good image data) then go for the upsampling route, there still might be something to gain with the extra 60 PPI. If that difference in print quality does not exist then I would take the downsampling route with Qimage. I avoid any up- or downsampling in Photoshop and try to keep the original image as it is in my archives.
Bottom line is still that question what actual visible resolution (and smoothness of gradations) a paper-ink-printer combination makes if fed with plenty of image data. After that you have to find the application with the best resampling algorithms, up and down. Then you can decide what to do, some print proofs may refine that process for the papers you use.
met vriendelijke groeten, Ernst
Try: http://groups.yahoo.com/group/Wide_Inkjet_Printers/
-
So you would never downsample unless the image's original resolution was above 720 dpi?
If my image started at 420 dpi, I've been downsampling to 360 dpi and sending it off to the Epson 3800 to print at 360 dpi native. My thinking was that it's better to lose 60 dpi of IQ from a PS-driven downsample than it is to lose 300 dpi of IQ from the printer driver's upsample interpolation, regardless of post-sampling sharpening.
And therein lies exactly the crux, the post-sampling sharpening. Upsampling to 720PPI will allow to do a better sharpening job than at half that sample density. Attached is an attempt to illustrate why I always upsample to the printer's highest native resolution, then sharpen for output.
I started with a crop from a Raw conversion (Capture One) without sharpening, and sharpened that with FocusMagic. Let's pretend it's printed on our screen at it's native resolution. Next I upsampled a copy of the unsharpened Raw conversion, resampled it to 2x it's native resolution, also used FocusMagic to sharpen (with 2x the radius, and a smidgen higher amount setting because the upsampled image was softer). I then downsampled the larger version to the same output size as the first version.
While the difference may be subtle to some, I know which one I prefer ..., the higher sampling density allowed better sharpening.
Cheers,
Bart
-
If there is a substantial difference visible in the print between the printer quality settings that ask for either 360 or 720 PPI rendering resolution (based on good image data) then go for the upsampling route, there still might be something to gain with the extra 60 PPI.
There is a benefit to outputting above 360 ppi data as 720 ppi data on Epson pro printers because the driver does a better dither of fine detail with Finest Detail on. So if you have enough resolution above 360, upsample to 720.
I've looked at upsampling below 360 ppi images to 720 ppi prior to output and there just isn't a free lunch, ya know? If the image is below 360, the best you can hope for is to upsample to 360 and then sharpen.
Personally, for ink jet printing I would NEVER downsample. Why waste your resolution?
There's actually a 1440 x 1440 mode in the x900 series printers (you can only access it when you select a proofing media). So, I'm thinking of testing images whose native rez is above 720 ppi and upsampling to 1440 ppi. I don't really think it's gonna do anything though.
-
Something needs clarifying: we are talking about raw conversions of digital camera files (e.g. in Lightroom, or Capture One) directly to the specific output required for Epson/Canon, with the re-sampling done by the raw converter, and then handing that off to the printer.
If that file already exists e.g. an image file where you don't have access to the RAW file, a scan from film, etc, then YMMV from what is being discussed here.
-
Personally, for ink jet printing I would NEVER downsample. Why waste your resolution?
One possible reason for down-sampling prior to printing would be to avoid aliasing artifacts from improper down-sampling. Photoshop's down-sampling has been criticized (Bart van der Wolf (http://bvdwolf.home.xs4all.nl/main/foto/down_sample/down_sample.htm)) and down-sizing by the printer (if it occurs on Epson printers--see previous thread discussion between Jeff and Bart) could pose similar problems. Lightroom reportedly has an improved downsizing algorithm, but if one is printing from Photoshop one could play it safe by doing his own downsizing prior to printing.
Regards,
Bill
-
Something needs clarifying: we are talking about raw conversions of digital camera files (e.g. in Lightroom, or Capture One) directly to the specific output required for Epson/Canon, with the re-sampling done by the raw converter, and then handing that off to the printer.
I'm talking about the native resolution of the file once the final print dimensions are determined...whether or not you resample in the raw converter or later in the pipeline, what's important is knowing the resolution once you are ready to print. Lightroom's Print module makes that very easy...it's a bit more difficult in Photoshop.
-
I don't have Lightroom, Jeff. What is the difference in the print modules?
I always prefer printing from Lightroom vs Photoshop for a variety of reasons–first off, Lightroom only cares about pixel dimensions, not size and resolution. So, if you need a small print, simply set the dimensions of the cell size and the resolution auto-flows...big print where you need less? again, it auto-flows. So you don't need to spawn off multiple files just to print different sizes.
Lightroom has an optimized upsampling that is image adaptive. If you upsample, LR interpolates between Bicubic and Bicubic Smoother depending on the size.
The last phase is output sharpening...which I had an involvement in since Adobe worked with PixelGenius to bring output sharpening to Lightroom.
The printing workflows is, in my opinion (which of course, I'm biased about) superior to Photoshop.
-
The printing workflows is, in my opinion (which of course, I'm biased about) superior to Photoshop.
Having done quite a bit of printing now from both Photoshop and Lightroom, I emphatically agree. And my only connection with either Adobe or Pixel Genius is that I am a happy customer of both.
Eric
-
Hello,
I must confess that I was still a bit skeptical about the advantages to up-sample to 300 or 600ppi (I use a HP Z3200) before printing, but reading Jeff and Ernst insisting (once again) that LR 3.x and Qimage algorithms are optimised to work that way I decided to give it a try. I own both softwares but usually print from Photoshop (or used to).
Today two clients cancelled their appointments giving me about to 3 hours free to play, so I carefully selected an image that was sharp, had various different textures and some low and high contrast areas. The image would be about 242ppi at 24" width. First I processed the image as I usually do and printed it 6 times on Cansosn Photo Satin Premium RC 270g paper on the following order:
1) Printed from PS without up-sampling at 242ppi
2) Printed from LR3.5 without up-sampling at 242ppi with "standard glossy" print sharpening
3) Printed from LR3.5 up-sampling to 300ppi with "standard glossy" print sharpening
4) Printed from LR3.5 up-sampling to 600ppi with "standard glossy" print sharpening, printer set to "maximum quality" (I know I should do that only if the image was originally above 300ppi, but I did just for fun)
5) Printed from Qimage up-sampling to 300ppi with "standard" print sharpening, "fusion" interpolation
6) Printed from Qimage up-sampling to 600ppi with "standard" print sharpening, "fusion" interpolation, printer set to "maximum quality" (again just for fun)
Then I numbered the prints and spent about an hour looking at them side by side on the light booth. I also showed them to my wife (and business parter) and some other photographers. Everyone agreed that (1), (2) and (4) were the worse in therms of sharpness and detail, proving that Jeff Schewe is right (as usual). But, curiously, there was no consensus about images (3), (5) and (6) and honestly I would approve any of those if presented separately. Image (6) was the anomaly, would not dare to say it is sharper and has more than (3) and (5), but it for sure is no worse.
I will try to test some other image tomorrow, this time an image that is above 300ppi.
Many thanks Jeff and Ernst, as I am now convinced!
Best regards.
-
Personally, for ink jet printing I would NEVER downsample. Why waste your resolution?
Enough textured matte art media around that does not give the image quality to make that 60 PPI + sharpening step visible. With several drivers the highest quality setting is not even available for media below that quality. There are good arguments to keep for example droplet size bigger and by that print resolution lower on media that is not at the highest quality standards. Dotgain of smaller droplets is higher than dotgain of small droplets, addressing larger droplets is easier than addressing smaller droplets, Dmax on matte media often is best with larger droplets. An Epson running at 720 PPI or 1440 PPI uses only the minimum droplets, excellent for very even, high quality surfaces but not always for rougher textures. In some cases users have to run at the highest quality and slowest printing speeds as the printer is not consistent in its output; more banding issues at the faster settings. This has nothing to do with the better quality of the transferred printer data but is a separate issue.
Then there is a practical issue: an image file that does not go through the upsampling and sharpening routines of for example Qimage gives a much lower amount of printer data to transfer. It is the opposite of lossy JPEG compression of images when you improve the image data with Q's routines; more information has to be transferred to the printer as there is more "quality" created. Sending 720 PPI data to the driver is already a lot more than sending 360 PPI data, better upsampling + sharpening routines expand that even more. With larger prints that can be problematic with memory and in any case both processing speed and printing speed suffer considerably when you always go for the 720 PPI + best upsampling and sharpening route. With 16 bit printer data that is again doubled. If the print quality gain is not visible or not worth it I would use 360 PPI + best resampling and sharpening. One of the advantages of the HP and Canon wide format printers is that the two rendering resolution are less demanding for printer data processing, choices are 300 and 600 PPI, the desktop models like the B9180 however have the 1200 PPI rendering print quality setting but produce smaller print sizes. Older Z3100 firmware had a 1200 PPI rendering printer quality setting while the newer Z3200 did not go higher than 600 PPI rendering resolution, I could not see a quality difference on detail between both. HP probably neither as the the latest Z3100 firmware no longer gives that 1200 PPI rendering choice. They could have improved the dithering/weaving in the same upgrade but the driver asks for 600 PPI input now.
I am not against that rule of upsampling whenever possible to the first rendering resolution above the available image resolution at print size but there is a nuance to that rule, as there is always a nuance. I have often used that upsampling rule with the notion that drivers ask for specific rendering resolutions, Qimage makes the user very aware of the parameters in the process. It also has the flexibility to create an optimal alternative: adaptable anti-aliased downsampling + smart print sharpening and the driver not interfering afterwards.
met vriendelijke groeten, Ernst
New: Spectral plots of +250 inkjet papers:
http://www.pigment-print.com/spectralplots/spectrumviz_1.htm
-
I have to print many huge images from Phase One P65+ files: These are large prints - over 75" long, so viewing will be relative, but results so far are very decent.
The print size was chosen to be 120ppi native - an exact 1/3 of 360dpi that the Epson requires. So we are not dealing with random uprez in awkward fractions here. But are people suggesting I should uprez it to 180dpi, or 360dpi (!) from RAW in C1 or from TIFF in PS, then sharpen, then print? It's hard to imagine this will make a lot of difference, when I have an exact multiple already, but opinion here seems to say it might.
The other option is QImage, which I know and like, allowing that do the work, BUT the images require special local area sharpening, so I can't use QI own sharpening procedures for that, sadly.
ps: I can't use Lightroom, as its Phase One profiles are not that good.
-
The print size was chosen to be 120ppi native - an exact 1/3 of 360dpi that the Epson requires. So we are not dealing with random uprez in awkward fractions here. But are people suggesting I should uprez it to 180dpi, or 360dpi (!) from RAW in C1 or from TIFF in PS, then sharpen, then print? It's hard to imagine this will make a lot of difference, when I have an exact multiple already, but opinion here seems to say it might.
While it can at times be beneficial to resample in integer fractional sizes compared to the printer's native resolution, I've seen no solid evidence that it really matters (it won't hurt either). It would matter if the printer driver has built in optimizations for certain fixed multiple interpolation/decimation ratios. My main hesitation in recommending e.g. 120 or 180 PPI is with the unknown quality of the interpolation to whatever the printer requires.
The other option is QImage, which I know and like, allowing that do the work, BUT the images require special local area sharpening, so I can't use QI own sharpening procedures for that, sadly.
You can do that, or perhaps even better. Qimage allows to File|Print To|File (e.g. in the output colorspace if you like). Then edit the print-file by using whatever latest invention you may have available for sharpening. In addition, you can have Qimage automatically interpolate to whatever resolution the printer driver feedback reports as optimal. Sharpening at that output resolution is also optimal, and there is no need to guess if/which integer fractions offer a benefit or not.
Cheers,
Bart
-
In addition, you can have Qimage automatically interpolate to whatever resolution the printer driver feedback reports as optimal. Sharpening at that output resolution is also optimal, and there is no need to guess if/which integer fractions offer a benefit or not.
Cheers,
Bart
thanks Bart. very useful. I understand all of this, except the bit about 'printer driver feedback' - where does the printer driver tell you what is wants? - or are you referring to the feedback on these pages telling us to work at 360dpi for Epsons.
I guess I should test at 720dpi output on the Epson too at this size. might be asking less from the interpolation, and end with different results.
-
thanks Bart. very useful. I understand all of this, except the bit about 'printer driver feedback' - where does the printer driver tell you what is wants? - or are you referring to the feedback on these pages telling us to work at 360dpi for Epsons.
Qimage takes away all the quesswork because it interrogates the printer driver about which PPI it expects based on the driver settings (e.g. paper type and ink choices). The printer driver responds with the optimal PPI (shown at the top right above the layout preview page) and Qimage automatically resamples the images to exactly that while observing the layout preferences, border widths, etc. It uses a superior resampling algorithm compared to the printer driver's to do the resampling, and can finally Smart sharpen based on the amount of interpolation from the original file dimensions.
If the resulting output size is such that the best quality would result in excessive printing times or spoolfiles, then one can instruct Qimage in the layout properties to drop the output resolution for Posters from Max, to High, or Medium, thus forcing a lower PPI resampling. You can also let Qimage cut up the images in multiple edge aligned posters of e.g. mounting panel size if the full print is larger than the available paper.
All of the print output can be directed at files (e.g. off-site printing or manual sharpening, retouching, reprinting, multiple printers, etc.) or directly at the printer.
I guess I should test at 720dpi output on the Epson too at this size. might be asking less from the interpolation, and end with different results.
Except for larger (spool) files, there may well be additional resolution to be had if not by the interpolation method used (which attempts to straighten jagged/stairstepped input and gives a smooth organic look to gradients and small detail). If 720 PPI output is warranted, then the printer driver should be set to finest detail, otherwise it will suggest 360 PPI maximum to Qimage.
Cheers,
Bart
-
Except for larger (spool) files, there may well be additional resolution to be had if not by the interpolation method used (which attempts to straighten jagged/stairstepped input and gives a smooth organic look to gradients and small detail). If 720 PPI output is warranted, then the printer driver should be set to finest detail, otherwise it will suggest 360 PPI maximum to Qimage.
Bart
Thanks Again Bart. very useful stuff. will re-familiarize myself with Qimage and try these.
One bit of confusion - my fault- is I meant setting the Epson printer to 720dpi, (rather than 1440 or 2880) for such very large prints, not the image ppi to 720. I'd imagine the printer asks less of Qimage, and so less interpolation is needed.
-
Thanks Again Bart. very useful stuff. will re-familiarize myself with Qimage and try these.
One bit of confusion - my fault- is I meant setting the Epson printer to 720dpi, (rather than 1440 or 2880) for such very large prints, not the image ppi to 720. I'd imagine the printer asks less of Qimage, and so less interpolation is needed.
The image PPI simply follows from the image pixels divided by the required output size. The number of required output pixels per unit length are determined by the printer driver settings. Qimage makes up for the difference by resampling to the exact number of pixels required, and do sharpening at the final output size. No intermediate files are required (everything happens in the background), although one can produce files for specific requirements.
If the output size results in only few original image pixels per output dimension (shown in the Queue list as image PPI), then it may not be necessary to use the full printer resolution capability. It won't technically hurt, but it may take a long time to print with increasingly little quality gain, in which case the Qimage output settings for resolution in the job properties can be relaxed a bit (Max->High->Medium). It's easier to do than to explain.
Cheers,
Bart
-
Has Epson improved their file resampling algorithms in the latest 9900 and 9880 driver models? I'm also wondering if the IPF8300 plug-in has major changes over Canon's previous model when it comes to pixel optimization and enlargement of files sent to it at smaller than ideal sizes.
j
-
While it can at times be beneficial to resample in integer fractional sizes compared to the printer's native resolution, I've seen no solid evidence that it really matters (it won't hurt either). It would matter if the printer driver has built in optimizations for certain fixed multiple interpolation/decimation ratios. My main hesitation in recommending e.g. 120 or 180 PPI is with the unknown quality of the interpolation to whatever the printer requires.
All image scaling introduce aliasing/imaging to varying degree. If the scaling factor is a division of two small numbers, the aliasing artifacts tends to be "consistent" for periodic input, and therefore less annoying.
-h
-
Has Epson improved their file resampling algorithms in the latest 9900 and 9880 driver models?
Resampling? Again I really don't think the driver resamples the image data....I think the print pipeline does. But if you are asking about the error diffusion dither of the x900 & x890 (not the x880 series) then yes, the dither is much finer.
-
If the resulting output size is such that the best quality would result in excessive printing times or spoolfiles, then one can instruct Qimage in the layout properties to drop the output resolution for Posters from Max, to High, or Medium, thus forcing a lower PPI resampling. You can also let Qimage cut up the images in multiple edge aligned posters of e.g. mounting panel size if the full print is larger than the available paper.
Cheers,
Bart
If the Max setting of resampling is not used in Qimage but High or Normal then the driver will do the last step of resampling to the rendering resolution in PPI of the print quality (dot resolution in dpi, more or less weaving strokes, etc) selected in the driver, for example from the 300 PPI delivered by Qimage to the 600 PPI asked for by the driver. That is what I understand of the Qimage choices on interpolation quality. At least that is obvious in the example mentioned as Qimage shows that lower output resolution next to choice while the higher requested rendering resolution is shown above the preview window.
met vriendelijke groeten, Ernst
Try: http://groups.yahoo.com/group/Wide_Inkjet_Printers/
-
Resampling? Again I really don't think the driver resamples the image data....I think the print pipeline does.
Somewhere between the application and the inkjet heads a resampling happens, up or down and to two or three specific rendering resolutions. The aliasing tests have shown that behaviour as discussed in the older thread. Possibly more number crunching is done internally on wide format printers than happens on desktop models in view of what is demanded and in view of printer costs. Where it happens after the application sends the data is not important I think, that specific rendering resolutions exist is however good to know.
met vriendelijke groeten, Ernst
New; 250+ Spectral plots soon extended with the Canon US catalog:
http://www.pigment-print.com/spectralplots/spectrumviz_1.htm
-
All image scaling introduce aliasing/imaging to varying degree. If the scaling factor is a division of two small numbers, the aliasing artifacts tends to be "consistent" for periodic input, and therefore less annoying.
Most real life images have a variety of detail and therefore offer various spatial frequencies. There's probably no frequency that doesn't align with the sampling period at some point and at the same moment most others don't align. Also, as for this thread's topic, producing large output usually means we're dealing with undersampled (for the output size) detail, thus forcing us to interpolate to get more pixels. That poses different challenges (filling in the 'gaps' with credible detail) than downsampling does.
Now, let's consider the seemingly simple case of 2x upsampling, say from 180 PPI input to 360 PPI printer requirement. That would mean that for each 2 pixels 1 pixel has to be interpolated exactly in between the two, and the two original pixels are unchanged. Sounds like a simple enough thing, fit for a printer driver which is built for speed.
Well, that may be an underestimation of the complexity we're faced with. Let me remind you that the original pixels, coming from an MF camera without proper AA-filter, are aliased to begin with (e.g. stairstepped edges and lines, especially when sharpened at native resolution, prior to upsampling).
We will possibly be better of by not leaving the original pixel values unchanged, but by prefiltering them in addition to adding interpolated pixels. What's more, the interpolation should ideally be done in linear gamma space, but we have no idea what gamma adjustments have taken place to create the image file we're sending to the printer driver (which therefore probably assumes nothing), and that's on top of the gamma 1/2.2 precompensation for display.
Cheers,
Bart
-
If the Max setting of resampling is not used in Qimage but High or Normal then the driver will do the last step of resampling to the rendering resolution in PPI of the print quality (dot resolution in dpi, more or less weaving strokes, etc) selected in the driver, for example from the 300 PPI delivered by Qimage to the 600 PPI asked for by the driver. That is what I understand of the Qimage choices on interpolation quality. At least that is obvious in the example mentioned as Qimage shows that lower output resolution next to choice while the higher requested rendering resolution is shown above the preview window.
Hi Ernst,
Yes, that what I understood as well, it lowers the quality/resolution of data sent to the printerdriver to save time (e.g. over a network) or to prevent spool memory issues. That doesn't mean that the driver doesn't still resample on it's own, if set to a normal or higher output resolution.
Cheers,
Bart
-
I always prefer printing from Lightroom vs Photoshop for a variety of reasons–first off, Lightroom only cares about pixel dimensions, not size and resolution. So, if you need a small print, simply set the dimensions of the cell size and the resolution auto-flows...big print where you need less? again, it auto-flows. So you don't need to spawn off multiple files just to print different sizes.
Lightroom has an optimized upsampling that is image adaptive. If you upsample, LR interpolates between Bicubic and Bicubic Smoother depending on the size.
The last phase is output sharpening...which I had an involvement in since Adobe worked with PixelGenius to bring output sharpening to Lightroom.
The printing workflows is, in my opinion (which of course, I'm biased about) superior to Photoshop.
Jeff, I do all my editing in Photoshop on drum-scanned large format B&W negatives and have always printed to my Epson 7900 from Photoshop to maintain a workflow that is reasonably straightforward. I'm trying to understand if there would be any significant value in changing my workflow to take my edited images and move them into LR for printing. I guess I'll have to run some tests myself, but you mention that LR has an "adaptive" upsampling that interpolates between Bicubic and Bicubic Smoother. If I were to choose to stick with PS for printing, what guidelines can you give for when to choose Bicubic vs. Bicubic Smoother when upsampling to the target 360 or 720?
Dave
-
If I were to choose to stick with PS for printing, what guidelines can you give for when to choose Bicubic vs. Bicubic Smoother when upsampling to the target 360 or 720?
For upsampling in Photoshop, use Bicubic Smoother (which produces the least artifacts), and then do the sharpening (e.g. Smart sharpening, More Accurate mode, try Lens blur first) at the final output resolution. You can push the Smart sharpening amount further in the absence of artifacts, and the radius would be your normal radius (0.7?) multiplied by the upsampling factor, for starters.
The result can be every bit as good as from Lightroom.
Cheers,
Bart
-
I understand the rule for an Epson printer that an image at or under 360 ppi, resample the picture to 360 ppi and then print.
General guide lines for up-sizing in Photoshop using bicubic smoother.
My question is how much can an image be up sampled before image quality will be effected?
Example: My original image size is 10.4 x 15.6 inches at 360 ppi. I print my large images at 20 x 30 inches. In this example I would have to double the width and height to stay at the same resolution (quadrupling the image size).
What I have been doing is resizing my image to 15.6 x 23.4 inches at 240 ppi and up sampling to 20 x 30 inches at 240 ppi and then printing. This method does not require up sampling by a large amount.
I do my output sharpening after resizing the print.
I do plan on printing an image both ways and testing.
-
My question is how much can an image be up sampled before image quality will be effected?
Pixel qualities differ, image content differs, media print qualities differ, print viewing distances can be taken into account, etc etc. In the end some proof strips and your eyes become the decisive factors. At some point you have to find a compromise between upsampling/sharpening artefacts and image softening, the best upsampling routines stretch that a bit. There are no strict rules in my opinion. The amount of upsampling can be better judged by setting the image size in Photoshop at the intended print size (without resampling), then check the PPI number there and compute the upsampling ratio with the knowledge that the printer/driver will ask for either 360 or 720 PPI.
met vriendelijke groeten, Ernst
New; 250+ Spectral plots soon extended with the Canon US catalog:
http://www.pigment-print.com/spectralplots/spectrumviz_1.htm
-
I will try to test some other image tomorrow, this time an image that is above 300ppi.
How did your second experiment work out?
-
Jeff, just for the sake of absolute clarity, are you talking about upsampling in the LR Print module using the Print Resolution option or by exporting the file to a new version and using the upsampling in the Export dialogue, then printing that file?
-
How did your second experiment work out?
I did not have the time to perform the test yet, unfortunately. But I will do it for sure and will let you know.
Best regards.
-
Jeff, just for the sake of absolute clarity, are you talking about upsampling in the LR Print module using the Print Resolution option...
Yes...that's one of the advantages of printing from Lightroom–you can check the native resolution with the Dimensions guide option and see what the resolution is at the print size and then decide what output resolution to use. This avoids having to spawn off multiple files for various print sizes...
-
Thanks, Jeff. That's what I figured. One of the nice efficiencies of LR. Just wanted to check to be sure I wasn't missing something.
-
How did your second experiment work out?
Today I managed to find some time to continue my experiment about resampling and printing from LR and QImage.
This time I found an image that would have 450ppi at my chosen print size, the exact middle between 300 and 600ppi (my printer is a HP Z3200). I printed it 5 times on the same paper as before (Canson Photo Satin Premium RC 270g), on the following order:
1) Printed without up-sampling or output sharpening at 450ppi
2) Printed from LR3.5 down-sampling to 300ppi with "standard glossy" print sharpening
3) Printed from LR3.5 up-sampling to 600ppi with "standard glossy" print sharpening, printer set to "maximum quality"
4) Printed from Qimage down-sampling to 300ppi with "standard - 5" print sharpening, "fusion" interpolation
5) Printed from Qimage up-sampling to 600ppi with "standard - 5" print sharpening, "fusion" interpolation, printer set to "maximum quality"
This time the game was more levelled. The image was a portrait with crisp sharp focus on the eyes, eyelashes and part of the hair and skin. The model had almost no make up, so de pores and the skin texture are quite evident on the crisp focus areas. At the same time the image also has a shallow depth of field and I was curious to see how the output sharpening routines would handle those areas, as one of the things I hate the most is sharpening artifacts on out-of-focus areas. Both QImage and LR 3.5 output sharpening hadled those areas very nicely without any image degradation.
The (1) was the worse, not bad actually, but below the others in terms of image detail and sharpness. All the other four images are excellent and almost identical to the naked eye. Looking very closely we can rule out number (2) also, but that is almost a stretch. The other three images (600ppi from LR and 300 & 600ppi from Qimaqe) are tied without hope of finding a winner.
This findings are somehow consistent with the opinions of some colleagues that is better to up-sample and apply output sharpening when printing from LR 3.5 or QImage. Again QImage produced an interesting anomaly, this time the down-sampled image being as good as the up-sampled one.
The next experiment will be the same thing on mat rag paper.
Best regards.
-
The next experiment will be the same thing on mat rag paper.
Yeah well, I would guess that a matte paper will top out with 300PPI with output sharpening...I doubt there's much to be expected or extracted from upsampling for matte/watercolor papers. in fact, some matte/watercolor papers won't even allow setting a max output (this is more specific to Epson, can't be sure about Canon/HP).
-
Yeah well, I would guess that a matte paper will top out with 300PPI with output sharpening...I doubt there's much to be expected or extracted from upsampling for matte/watercolor papers. in fact, some matte/watercolor papers won't even allow setting a max output (this is more specific to Epson, can't be sure about Canon/HP).
Sure Jeff, I tend to agree, but will try 600ppi also just to see what happens. I think HP allows the setting, but never actually tested it.
Will see!
-
Sure Jeff, I tend to agree, but will try 600ppi also just to see what happens. I think HP allows the setting, but never actually tested it.
Will see!
Hi Geraldo,
There is a benefit to be had from first upsampling to 600 PPI, doing good sharpening at that output resolution, and then print, even if the output medium won't resolve the finest detail. The benefit is that when sharpening the highest spatial frequencies (the ones that don't resolve due to the medium), the next lower spatial frequencies (the ones that do resolve on the medium) also get a boost but without the artifacts. When you know that the finest details cannot be resolved by the medium, then you can push the highest frequency sharpening a bit further than usual.
Cheers,
Bart
-
There is a benefit to be had from fist upsampling to 600 PPI, doing good sharpening at that output resolution, \
Hi Bart, just out of interest, which program are you using/recommending to upsample?
I have a good MF digital file to make a large print (about 147x190cm). Uninterpolated it is a sweet 120dpi exactly, upsampling this to 360dpi before applying my sharpening routine and sending to the printer, is what is suggested here, but PS does a so-so upsizing. What are people suggesting? I could use Qimage with 'output to file', and then do my special sharpening, but it balks at such large output, hitting 32bit memory limits, even with an 8bit file. So that's off the list.
-
Hi Bart, just out of interest, which program are you using/recommending to upsample?
I use different routes, different pograms, depending on the rest of the workflow.
I have a good MF digital file to make a large print (about 147x190cm). Uninterpolated it is a sweet 120dpi exactly, upsampling this to 360dpi before applying my sharpening routine and sending to the printer, is what is suggested here, but PS does a so-so upsizing.
Actually, Photoshop's Bicubic Smoother isn't all that bad, because it doesn't create too many artifacts which sharpening will only 'enhance' further. It does require, and can stand, some additional sharpening, Smart sharpening (Lens Blur, more accurate) at least. Better deconvolution sharpening would be ideal, but it might take a while to execute (processor intensive).
Another option is to use ImageMagick which by default uses the Mitchell Netravali type of reconstruction filter for upscaling, and it allows to optionally modify the settings depending on how sensitive the source data is for generating (e.g. ringing) artifacts. However, ImageMagick's command line driven interface isn't everybody's cup of tea.
Another option is to use Photozoom Pro which allows to boost sharp edges as rasterless vector data (a bit similar to the Blow-up plug-in) but less expensive. One needs to exercise restraint in applying the edge enhancement, because it can lead to a mental disconnect between the representation of edges and surface detail, which IMHO looks unnatural.
However, these may all may cause file size issues further down the chain, except Photozoom Pro. Photozoom Pro also works as a Photoshop plugin, so it could be used when one prints remotely direct from Photoshop, and PhotoZoom Pro 4 enlarges images up to 1 million by 1 million pixels.
I could use Qimage with 'output to file', and then do my special sharpening, but it balks at such large output, hitting 32bit memory limits, even with an 8bit file. So that's off the list.
I have understood that it's not a direct Qimage problem, but a potential problem with the TIFF library specification. TIFFs are limited to 4GB file sizes due to its 32-bit adressing method. Some libraries are written with signed integers, and are therefore limited to 2GB file sizes. The upcoming BigTIFF standard is not final yet. It would be a nice addition if Qimage could write Photoshop PSB files, but that would probably lock the output files to Photoshop only as printer driver, which wouldn't seem to benefit Mike Chaney much especially since it forces him to change his code when the PSD proprietary (?) format is changed, so I can understand his hesitation. A 64-bit Qimage version appears to be scheduled for later in 2012, but that is not necessarily coming with a large file output option.
Which leave us with printing directly from Qimage, and that should work fine given the fact that all conversions (colorspace and interpolation) are handled on a block by block basis, which are then immediately fed to the printspooler. Since Qimage also allows to tweak the amount of Smart Sharpening it applies after resampling, it seems like a good compromise. For remote printing it may not be possible to persuade them to use Qimage, which would lead us back to one of the earlier options mentioned.
Cheers,
Bart
-
Is it better to sharpen AFTER a QImage resample using QImage print sharpening than it is to sharpen with Photoshop (or PhotoKit sharpen)?
I'm hearing a lot of complaints that Qimage doesn't allow print sharpening with other tools besides Qimage, unless one first saves a copy of the file as a TIFF. But doing that contravenes the whole point of not having to spin off separate resampled files to the printer. What if someone needs more selective sharpening application than the QImage's Low, Med, High sharpening setting will permit?
Or is it better to use the QImage sharpening tool because it is more optimized for it's upsampling algorithm. In other words, will Photoshop sharpening monkey up the Qimage resample quality? If so, would this work........
First do all print sharpening in Photoshop. That is, over-sharpen.
Then, resample using Qimage,
Then, apply LOW Qimage sharpening (or no Qimage sharpening)
Or is it always best to print sharpen AFTER resampling?
-
Gerald, I'm not surprised that your experiment demonstrated that upsampling/downsampling to/from 600ppi/300ppi from a starting point of 450ppi gives insignificant improvement. This is likely because our eye does not resolve more detail beyond 300ppi. That is why your first experiment showed a considerable improvement while your second one didn't.
I'm curious why you didn't test Photoshop's bicubic resampling tool. Are you convinced it is inferior to Lightroom and Qimage for the range of up/down-sampling jobs you typically do?
-
This is likely because our eye does not resolve more detail beyond 300ppi.
Actually, that's not true...the exact resolution of the eye is based on the viewing distance, not a given PPI. So, if you are viewing an image from 8", a normal person can see about 428PPI. As you move back, the eye can resolve less so at 18" the eye can only resolve 191PPI. This is based upon Bruce Fraser's calculations and the eye's stated resolution of one minute of arc (or 1/60th of a degree). Of course, that's based on high contrast line pairs, a low contrast line pair (more like continuous tone photography) the resolving capability goes down.
So, what resolution you need is based on the final viewing distance...on the other hand, even with big prints photographers tend to walk into an image and look close–Bruce used to say the intended viewing distance of a photographer is limited by the length of his nose :~)
Low resolution is fine for big output that will NEVER bee seen up close but to have a really good output that stands close inspection, you need more resolution.