Luminous Landscape Forum

Raw & Post Processing, Printing => Printing: Printers, Papers and Inks => Topic started by: Asael on February 01, 2018, 10:07:00 pm

Title: Resizing up or down for printing, what is the tipping point?
Post by: Asael on February 01, 2018, 10:07:00 pm
I am using a Canon ipf8300. For images that are bellow 300ppi I resize to 300 in Photoshop prior to printing. For images that are above 600ppi I resize to 600. But what to do about images that are in between 300 and 600? Resize up to 600 or down to 300? I currently resize up to 600 if the original image is above 500, and resize down to 300 if the images is below 500, but I realize that my 500ppi rule is not based on any real experiments and am wondering what are the rules other people use to decide if to resize up or down. Thanks!
Title: Re: Resizing up or down for printing, what is the tipping point?
Post by: Schewe on February 02, 2018, 01:02:01 am
Below 300 PPI, upsample in Photoshop using Preserve Details in the resample and then do output sharpening after upsampling. If the image is above 300 PPI but below 600 PPI, upsample to 600 PPI and sharpen. If you image is above 600 psi, there may, depending on your printer/driver be able to upsample to 1200 PPI and see benefit. Somebody with better experience with Canon's might chime in...

But, if there isn't a driver option to make use of 1200 PPI then yes if the image is over 600 PPI you would want to downsample in Photoshop and then output sharpen.

For Epson the magic numbers are 360/720 PPI.
Title: Re: Resizing up or down for printing, what is the tipping point?
Post by: Doug Gray on February 02, 2018, 01:07:17 am
Below 300 PPI, upsample in Photoshop using Preserve Details in the resample and then do output sharpening after upsampling. If the image is above 300 PPI but below 600 PPI, upsample to 600 PPI and sharpen. If you image is above 600 psi, there may, depending on your printer/driver be able to upsample to 1200 PPI and see benefit. Somebody with better experience with Canon's might chime in...

But, if there isn't a driver option to make use of 1200 PPI then yes if the image is over 600 PPI you would want to downsample in Photoshop and then output sharpen.

For Epson the magic numbers are 360/720 PPI.
I have tested this on the Canon 9500 II. It uses the nearest neighbor when resampling so it's best to up or down resample directly to 600 before printing.
Title: Re: Resizing up or down for printing, what is the tipping point?
Post by: tom b on February 02, 2018, 01:37:33 am
Follow Schewe's advice. However, take notice that no one buying your photos will notice the difference between the two.

Best luck in your endeavours,
Title: Re: Resizing up or down for printing, what is the tipping point?
Post by: Ernst Dinkla on February 02, 2018, 05:45:16 am
Follow Schewe's advice. However, take notice that no one buying your photos will notice the difference between the two.

Best luck in your endeavours,

Well you may see a nice moiré appearing in your prints if you rely on the downsampling of the printer driver loading a 500 PPI image at 1:1 print scale and having print quality setting ask for a 300 PPI input. Depends on the subject but fabrics can already cause that on glossy papers. In general if the paper/printer combination can yield a better print quality with that 600/720 PPI or higher input request then use it.

On the other hand with very large prints and always going for upsampling, to 600/720 or beyond, the time on processing and printing becomes very long and/or the process may go belly up in that process. Then going for a lower print quality is often the only choice remaining, I do not rely on the downsampling of the driver then but rely on the anti-aliased downsampling of Qimage Ultimate, not available in all programs. That anti-aliased downsampling in QU is also a slow process, actually going for upsampling where possible goes faster. In practice it means that I try going up all the time but when it fails I select the path down but with precaution.

Met vriendelijke groet, Ernst

http://www.pigment-print.com/spectralplots/spectrumviz_1.htm
March 2017 update, 750+ inkjet media white spectral plots
Title: Re: Resizing up or down for printing, what is the tipping point?
Post by: Bart_van_der_Wolf on February 02, 2018, 06:34:58 am
In practice it means that I try going up all the time but when it fails I select the path down but with precaution.

Yes, that's my rule of thumb as well, but Qimage Ultimate takes away most of that decisionmaking and automatically does the resampling and subsequent output sharpening. It also does not require to do a prior specific resampling for a given output size, it manages that on-the-fly. Especially useful if multiple print sizes are required.

Also nice is QU's capability to print quasi 1200 PPI in 'Overdrive' mode, which comes in handy if e.g. a smaller print is needed from a large image file, such as a stitched panorama. The improvement in image detail, while subtle at times, is visible. So in case of >600 PPI image detail, using 'Overdrive' mode can be beneficial compared to downsampled to 600 PPI output.

Upsampling, instead of tossing away excess detail, always offers the benefits for good output sharpening because there is more real detail to work with.

Cheers,
Bart
Title: Re: Resizing up or down for printing, what is the tipping point?
Post by: deanwork on February 02, 2018, 09:21:06 am

I've found QImage works very well in regard to upsampling and I use it on my HP z printers with a pc.

But with my Canon large format, which is a faster printer I don't use Q Image, I use the Canon plugin for Photoshop which I have found to work very well in regard to output sharpening. I'm using a Mac with the Canon and Epson printers.

For me, whether to upsample a file or not depends on the size of the file, what the quality and character of the file is,and  often even what kind of camera was used.

A lot of files I see when large prints are desired, just can't handle upsizing to 300 ppi, or 360 ppi without fringing badly that may require a lot of post retouching if they are even fixable that way. In this world we have people wanting large  exhibition prints from jpegs, iPhone jpeg files, and just generally files that are too small. I end up having to drop the ppi to 240, 180 or even 150 to make big prints from small files. But I'm using the Canon Plugin not printing from Lightroom or qimage with this 8300 printer.

John







Yes, that's my rule of thumb as well, but Qimage Ultimate takes away most of that decisionmaking and automatically does the resampling and subsequent output sharpening. It also does not require to do a prior specific resampling for a given output size, it manages that on-the-fly. Especially useful if multiple print sizes are required.

Also nice is QU's capability to print quasi 1200 PPI in 'Overdrive' mode, which comes in handy if e.g. a smaller print is needed from a large image file, such as a stitched panorama. The improvement in image detail, while subtle at times, is visible. So in case of >600 PPI image detail, using 'Overdrive' mode can be beneficial compared to downsampled to 600 PPI output.

Upsampling, instead of tossing away excess detail, always offers the benefits for good output sharpening because there is more real detail to work with.

Cheers,
Bart
Title: Re: Resizing up or down for printing, what is the tipping point?
Post by: NAwlins_Contrarian on February 02, 2018, 09:52:09 am
Just remember that all of this applies mostly to bordered printing, not borderless.

The discussion of resampling to a printer resolution is largely premised on Lightroom or Qimage or whatever doing higher-quality resampling than the printer or its driver would. If you want to make sure that the printer does not resample, then you probably have to print in bordered more.

If you print in bordered mode, you should be able to resample in Lightroom or Qimage or whatever to the exact pixel count for the output size and resolution you're using. E.g., if I print an 8x10 inch image area on a letter-size (8.5x11 in) sheet with the Canon Pro-100, I resampling to 2400x3000 pixels means the printer won't do any resampling.

However, if you print borderless, the printer is going to insist on some overspray, for which it will resample unless you happen to have sent it the exactly-right number of pixels. With some printers the degree of overspray is fixed--which is not to say that it is readily and precisely discernible. With others it's adjustable. For example, the Pro-100 has a four-position overspray control (called "extension"), and based on my measurements, it appears to me that those settings correspond to about 0.06%, 1.1%, 1.6%, and 2.6%. So if I want to print an 8x10 inch picture using (instead of letter-size) 8x10 paper, depending on settings, the printer appears to need, and therefore if necessary resample to, roughly 2401x3002, 2425x3032, 2439x3049, or 2463x3079. Unfortunately, I don't know how to determine precisely.
Title: Re: Resizing up or down for printing, what is the tipping point?
Post by: datro on February 02, 2018, 10:58:22 am
Yes, that's my rule of thumb as well, but Qimage Ultimate takes away most of that decisionmaking and automatically does the resampling and subsequent output sharpening. It also does not require to do a prior specific resampling for a given output size, it manages that on-the-fly. Especially useful if multiple print sizes are required.

Cheers,
Bart

The "downsampling" part of this discussion is VERY timely for me, as I have recently been running some tests to compare Qimage downsampling + sharpening vs my own custom script that is largely based on your previous Imagemagick work from several years back (though my script runs with IM7, is "tuned" for grayscale images, and allows the user to provide desired print size width and the desired resolution to calculate the geometry parameter for the resize).

My overall workflow (which drives my downsampling research) is as follows:

1. Scan 4x5 B&W negatives on my drum scanner at 4000dpi
2. Process in Photoshop (without any resizing) to taste, save a "master" TIFF
3. Downsize to my desired print size at 720 dpi and output sharpen
4. Print using QTR and Piezography Pro inks

Step "3" is what I'm trying to optimize.  The requirement for 720 dpi is driven by the fact that QTR (according to Roy Harrington) always wants a 720 dpi input file.  If it gets anything else, it will do its own resizing and I want to avoid that (no different than if I was printing using the standard Epson driver).

While Qimage does a credible job with the "print to file" functionality, it is a bit cumbersome for that purpose and I wanted something more streamlined for use with QTR as my printing engine, hence my effort to write my own script using Imagemagick.  It is already clear to me that either Qimage or the Imagemagick approach can do a better job than using the Photoshop "Bicubic Sharper" algorithm, especially since I can control the amount of sharpening on the resampled data to suit the image and the media I'm using.

In your experience, are there other downsampling + sharpening tools I should be looking at as well?  Do you feel Qimage is capable of matching or improving on the Imagemagick approach?

Thanks,
Dave






Title: Re: Resizing up or down for printing, what is the tipping point?
Post by: Binartem on February 02, 2018, 11:06:49 am
As a general rule, if the original is under 600 PPI at the given print size, always upsample to 600.  Downsampling to 300 introduces an extra step (anti-aliasing) and as with most things, we want to go with the simplest methods (fewer image manipulations).  For Epsons, use 720 PPI.  Above 600/720, it's more of a user preference because above that, it takes close scrutiny to discern differences.  Some Canon drivers actually allow you to run the driver at 1200 PPI if you turn off the "prevention of data loss" feature.  Whether you need that level of detail depends highly on the nature of what you are doing.  For most normal size prints, viewers won't notice a difference in detail between 600 and 1200 PPI at normal viewing distances.

Regards
Title: Re: Resizing up or down for printing, what is the tipping point?
Post by: Asael on February 02, 2018, 02:27:32 pm
Follow Schewe's advice. However, take notice that no one buying your photos will notice the difference between the two.

Best luck in your endeavours,

While it maybe true that most of my buyers would not notice the difference, but I (using a magnifying glass) may, so it's worth the effort for me :)
Title: Re: Resizing up or down for printing, what is the tipping point?
Post by: Asael on February 02, 2018, 02:34:28 pm
I experimented with the Canon plug-in printing for PS a few years back, and could not see a difference in the output between it and PS.
I may give Qimage a try.
Thanks all.
Title: Re: Resizing up or down for printing, what is the tipping point?
Post by: deanwork on February 02, 2018, 03:01:25 pm

How that functions depends on how you have tweaked the output sharpening slider.



I experimented with the Canon plug-in printing for PS a few years back, and could not see a difference in the output between it and PS.
I may give Qimage a try.
Thanks all.
Title: Re: Resizing up or down for printing, what is the tipping point?
Post by: Bart_van_der_Wolf on February 02, 2018, 07:42:21 pm
My overall workflow (which drives my downsampling research) is as follows:

1. Scan 4x5 B&W negatives on my drum scanner at 4000dpi
2. Process in Photoshop (without any resizing) to taste, save a "master" TIFF
3. Downsize to my desired print size at 720 dpi and output sharpen
4. Print using QTR and Piezography Pro inks

Step "3" is what I'm trying to optimize.  The requirement for 720 dpi is driven by the fact that QTR (according to Roy Harrington) always wants a 720 dpi input file.  If it gets anything else, it will do its own resizing and I want to avoid that (no different than if I was printing using the standard Epson driver).

While Qimage does a credible job with the "print to file" functionality, it is a bit cumbersome for that purpose and I wanted something more streamlined for use with QTR as my printing engine, hence my effort to write my own script using Imagemagick.

Hi Dave,

I agree, since you will not be printing with Qimage, that its exclusive use for resampling and Output sharpening does not produce the best workflow. Qimage would mainly add more benefits when it's used to print with. While Qimage adds several other things, like dithering, nesting, halo-free output sharpening after resampling, the use of templates, etc., a much slower and more dedicated process by using ImageMagick can produce somewhat higher quality (because it can be tuned for a specific scenario like yours).

Quote
It is already clear to me that either Qimage or the ImageMagick approach can do a better job than using the Photoshop "Bicubic Sharper" algorithm, especially since I can control the amount of sharpening on the resampled data to suit the image and the media I'm using.

IMHO, Bicubic Sharper is not very good, and it's not tunable, so your finding doesn't come as a surprise.

Quote
In your experience, are there other downsampling + sharpening tools I should be looking at as well?  Do you feel Qimage is capable of matching or improving on the Imagemagick approach?

As I said, because ImageMagick is slower (because it handles 16-bit/channel data, floating point number calculations, a 2D filter-kernel convolution sharpening, and a scripted sequence of many events), and because of how the image resampling is done, the result can be a bit better. Hard to improve that.

Output sharpening could be tuned a bit better in ImageMagick for the specific workflow you use (with predictable amount of downsampling and probably predictable output media, and thus output sharpening), or you could use a Photoshop plugin like FocusMagic that uses deconvolution sharpening and can be adjusted for dealing with film scans instead of digital captures, and can the result can be improved a bit more if used as a Luminosity sharpening layer with Blend-if to avoid clipping. But that means you need to use Photoshop again after ImageMagick has done the downsampling in step 3. Not an ideal workflow, but doable. Afterall, scanning is not very fast either.

Cheers,
Bart
Title: Re: Resizing up or down for printing, what is the tipping point?
Post by: Stephen Ray on February 02, 2018, 10:01:27 pm
It would be super if someone could explain how an individual RGB color pixel @ 360 ppi from an image file gets translated to a matrix of printer dots @ 1440 of that same, single RGB color pixel.

What would be super super is for someone to include a macro photo or scan of the result for demonstration.

Because most true RIPs have some sort of setting to compensate for print length being too short or too long, what then happens to any pixel-for-pixel reproduction concepts? 

Title: Re: Resizing up or down for printing, what is the tipping point?
Post by: NAwlins_Contrarian on February 02, 2018, 11:29:24 pm
Quote
It would be super if someone could explain how an individual RGB color pixel @ 360 ppi from an image file gets translated to a matrix of printer dots @ 1440 of that same, single RGB color pixel. What would be super super is for someone to include a macro photo or scan of the result for demonstration.

Unfortunately I don't have the perfect reference or succinct explanation to pop in here. Just as a basic example, a CMYK printer simulates red by printing some mix of magenta ink dots, yellow ink dots, black ink dots (to make it darker), and white space (no ink, to make it lighter). So maybe the printer simulates a mid-tone red with eight magenta dots and eight yellow dots per red pixel; and a dark red with five magenta dots, five yellow dots, and six black dots per pixel. But to get a general idea of how a limited number of colors of ink dots are used to simulate millions of colors of pixels, here is a pretty good article on the semi-historical process that will give you some ideas how photo inkjets might do it--although photo inkjets are arguably more sophisticated and higher quality:
https://en.wikipedia.org/wiki/Halftone (https://en.wikipedia.org/wiki/Halftone)

Thread contributor (and printing guru) Bruce Schewe reproduced an interesting section from his book The Digital Print, complete with enlarged examples showing to some extent ink colors forming pixels (albeit not enlarged that much, and done for another purpose) here:
https://www.digitalphotopro.com/technique/photography-workflow/the-right-resolution/ (https://www.digitalphotopro.com/technique/photography-workflow/the-right-resolution/)

Some additional background that illustrates some of the general concepts:
https://en.wikipedia.org/wiki/Dither (https://en.wikipedia.org/wiki/Dither)

Also:

Quote
Some Canon drivers actually allow you to run the driver at 1200 PPI if you turn off the "prevention of data loss" feature. Whether you need that level of detail depends highly on the nature of what you are doing.  For most normal size prints, viewers won't notice a difference in detail between 600 and 1200 PPI at normal viewing distances.

Although I don't know which Canon drivers those are, it seems to me a significant point that the more pro-oriented Canon photo printers--the sort more likely to have these sorts of advanced features--are mostly 2400x1200 dpi. If you print a 1200 ppi image with a 2400x1200 dpi printer, you only have two ink-spots per pixel. I seriously doubt that the printer can do a decent job accurately and smoothly simulating millions of colors with only two dot-spots and eleven or fewer colors of ink. And if you mitigate that with pixel-level dithering, you've defeated the benefit of going to 1200 ppi anyway.

Title: Re: Resizing up or down for printing, what is the tipping point?
Post by: Schewe on February 03, 2018, 01:26:27 am
Thread contributor (and printing guru) Bruce Schewe reproduced an interesting section from his book The Digital Print, complete with enlarged examples showing to some extent ink colors forming pixels (albeit not enlarged that much, and done for another purpose) here:
https://www.digitalphotopro.com/technique/photography-workflow/the-right-resolution/ (https://www.digitalphotopro.com/technique/photography-workflow/the-right-resolution/)

While I got a bit of a laugh to be renamed Bruce, my name is Jeff...however, I did learn a lot from a friend (and biz partner) named Bruce so in a way, I do have a bit of Bruce in me :~)

As to the conversion of pixels/inch into droplets/inch, that's what the inkjet printer does in producing error diffusion or some other exotic dithering to make images in ink on paper. see: Floyd–Steinberg dithering (https://en.wikipedia.org/wiki/Floyd–Steinberg_dithering), Error diffusion (https://en.wikipedia.org/wiki/Error_diffusion), or just do a Google search for error diffusion dithering.
Title: Re: Resizing up or down for printing, what is the tipping point?
Post by: texshooter on February 03, 2018, 03:43:01 am
But, if there isn't a driver option to make use of 1200 PPI then yes if the image is over 600 PPI you would want to downsample in Photoshop and then output sharpen.

I could have sworn you once said  never to downsample for print because "throwing pixels away" would be more degradative?

Perhaps you had actually said not to downsample the original file without making a copy first.  I've never printed from Lightroom, which spins off iterations without altering the original, so I must've heard you wrong.
Title: Re: Resizing up or down for printing, what is the tipping point?
Post by: deanwork on February 03, 2018, 08:41:39 am

I remember in one of his books what Bruce Fraser said about upsampling. He said I only have "one thing to say about upsampling, don't do it."

Of course technology changes as do workflows.





I could have sworn you once said  never to downsample for print because "throwing pixels away" would be more degradative?

Perhaps you had actually said not to downsample the original file without making a copy first.  I've never printed from Lightroom, which spins off iterations without altering the original, so I must've heard you wrong.
Title: Re: Resizing up or down for printing, what is the tipping point?
Post by: Bart_van_der_Wolf on February 03, 2018, 09:09:55 am
I could have sworn you once said  never to downsample for print because "throwing pixels away" would be more degradative?

Which would be correct, unless you have an excess of data points that the printer driver cannot handle, or which causes driver buffer-memory issues. In that case it's better to take control in order to optimize the down-sampling quality and output sharpening.

Quote
Perhaps you had actually said not to downsample the original file without making a copy first.  I've never printed from Lightroom, which spins off iterations without altering the original, so I must've heard you wrong.

That's the issue. Either the application one prints with does a good job with on-the-fly conversion to the native printer resolution, or we are at the mercy of poor driver resampling without output sharpening.

Cheers,
Bart
Title: Re: Resizing up or down for printing, what is the tipping point?
Post by: Bart_van_der_Wolf on February 03, 2018, 09:19:26 am
But, if there isn't a driver option to make use of 1200 PPI then yes if the image is over 600 PPI you would want to downsample in Photoshop and then output sharpen.

I agree. It's better to have control over the conversion, and add output-sharpening to compensate for resampling losses and specific media features (ink diffusion or contrast).

Quote
For Epson the magic numbers are 360/720 PPI.

In general, that's correct. However, there are recent Epson models that seem to be using 300/600 PPI, just like Canon and HP. An application like Qimage would detect (and use) that automatically because it interrogates the printer driver for those details.

Cheers,
Bart
Title: Re: Resizing up or down for printing, what is the tipping point?
Post by: NAwlins_Contrarian on February 03, 2018, 11:04:01 am
Quote
While I got a bit of a laugh to be renamed Bruce, my name is Jeff.

Doh! First, please accept my apology for making that mistake.

Second, just in case you were wondering, Schewe is not a common name around here, I only know one, and his name is Bruce. Apparently my end-of-the-day brain wasn't up to keeping it straight.
Title: Re: Resizing up or down for printing, what is the tipping point?
Post by: datro on February 03, 2018, 05:06:03 pm
... Afterall, scanning is not very fast either.

Cheers,
Bart

Hi Bart,
Yes, you are quite right  :).  Scanning a 4x5 negative on my Howtek 4500 @4000dpi takes about 1hour 10 minutes.  But the results are worth it, and I can do other work in parallel so it is not wasted time.

Thanks very much for your thoughts and suggestions (and also the previous work on your ImageMagick script!).  Much appreciated.

Dave
Title: Re: Resizing up or down for printing, what is the tipping point?
Post by: Doug Gray on February 03, 2018, 10:19:05 pm
As part of resampling, the diffusion dithering, more so than the ink dot size, that occurs spreads the location of a printed pixel a bit. This creates a low pass filtering effect similar to the "zero hold" low pass filtering that occurs with a DTOA output is held constant between clocks.  On the "zero hold" DTOA it produces about a 4 dB drop at Nyquist. The dithering, or pixel spreading, is a somewhat larger effect that occurs but can also be measured.

I ran an experiment with Canson Rag Photo on my 9500II and there is about a 5 dB drop off at 250 PPI and 2 dB at 150 PPI when printing 600 PPI images with the finest setting. It seems possible that this characteristic could be measured for a given printer and an appropriate sharpening filter applied. There are limits to such a filter. Pixels that would be pushed beyond the media white or black would have to be spread out so one couldn't completely compensate though I believe that rare in most real photos.

I'm also curious whether glossy is better.
Title: Re: Resizing up or down for printing, what is the tipping point?
Post by: tom b on February 03, 2018, 11:26:52 pm
Michael did a test with prints from two widely different cameras. The results were very interesting

Please, do two prints (or more), ask people if you can tell the difference between the two and report back. I know what the results should be, but I'm willing to be surprised.

Cheers,
Title: Re: Resizing up or down for printing, what is the tipping point?
Post by: unesco on February 04, 2018, 03:00:39 am
Yes, moire is often visible when high res image is downsampled by Epson driver. When I saw it the first time I could not find a reason and it have taken me a while to locate the problem.

As for QTR (up/down)sampling - you can try using a file just you have since QTR does spline sampling process to 720 ppi which works fine.
Title: Re: Resizing up or down for printing, what is the tipping point?
Post by: texshooter on February 11, 2018, 08:37:12 pm
Interesting twist.  Vincent Versace says you should print at 270ppi, which is 90ppi lower than the printer's native 360ppi resolution (Epson), when

1. You have a bronzing problem, or
2.  When using papers with high dot gain absorption (eg, Epson Somerset Velvet)

video,  start 32:15 (https://m.youtube.com/watch?v=77sx6lvXTac)
Title: Re: Resizing up or down for printing, what is the tipping point?
Post by: Schewe on February 12, 2018, 04:53:36 pm
Vincent Versace says you should print at 270ppi, which is 90ppi lower than the printer's native 360ppi resolution (Epson)...

Vinnie is wrong.