One of the most controversial topics in inkjet printing is whether to resize and resample an image for printing. For those not familiar with this, let me explain.  Most printer drivers have a “native” resolution that they expect images in.  Canon and Epson are different, with Canon expecting 300 or 600ppi, and Epson uses 360 or 720ppi.  (Note that I’m using PPI, or pixels per inch, as it’s technically more correct than DPI in this context.)  What if your digital image is not large enough, i.e. doesn’t have enough pixels to print the physical size you want at the applicable resolution?  Should you use Photoshop to resize / resample it up to that resolution?

This issue has been debated with the fervour of a religious war.  Search the Luminous Landscape (LuLa) printing forum for example.  I have no doubt that there are similar threads on DPReview and there’s one on The Online Photographer as well.  A lot of the discussion involves the behaviour of the Epson driver and the operating system and how they interact.  For Piezography the issue is how QuadToneRIP (QTR) and the operating system interact.

Roy Harrington, the creator of QTR, and Jon Cone, creator of Piezography, both suggest to simply give QTR whatever optical resolution you have and don’t bother resampling.  So rather than resampling an image to 720ppi, just set the physical size to whatever size you want to print, say A4 or A3 or A3+, but without resampling, and let the pixels per inch adjust.  This may mean sending an image to print at 250.50ppi, for example.  However a closer examination reveals that Roy and Jon are saying this for quite different reasons.

The definitive word from Roy was this post on the Yahoo QTR group in which he says “QTR works on 720ppi input data, so on Mac OS X that is provided by the system, on [a] PC there is an extra step to resample to 720ppi for QTR.”  This is quite a short statement.  Expanding it a little, what I believe he is saying is that QTR expects 720ppi.  If it doesn’t get it on Windows, then QTR will do the resizing / resampling.  However on Mac OS X, which I rarely use, it’s my understanding that the resizing is done by the printing pipeline rather than QTR.

Jon Cone, on the other hand, has advocated just giving QTR whatever optical resolution you have because (he says) QTR will use whatever you give it, be it less than 720ppi, like 250.5ppi, or more, say 1000ppi.  He claims that there’s no point in resampling up to 720ppi.  Moreover, the 30 April 2015 IJM newsletter discussed the resolving power of Piezography and said: “Piezography K7 curves can resolve over 1000 pixels per inch“, a statement repeated in the New Piezography Manual on p68.

I can’t reconcile what Jon says with what Roy has said.  If QTR (or the Mac OS X printing pipeline) will resample up to 720ppi, then I can’t see the harm in searching for a better way of doing it, even if the benefit is negligible.  And I can’t see the point of sending 1000ppi to QTR when it’s only going to be downsized.  I haven’t tested downsizing to 720ppi myself, but given that downsizing softens an image, I’d have thought that it would be better to find the best combination of downsizing to 720ppi and sharpening yourself, rather than leave it to QTR.  Although in most instances the difference in the print by doing so is going to be marginal at best, as Roy suggested.

The matter remains unresolved, but Roy wrote the program.

Using Windows, I’ve recently compared one print where I sent a file to at 250.50ppi and let QTR resample, and another where I resampled to 720ppi using Qimage and its high quality upsampling.  The image resampled using Qimage was slightly better, but I had to look really carefully to see the difference.  Which supports what Roy goes on to say in that post: “You might create artificial files that can demonstrate these re-sampling but for real images I really don’t think any of this matters a whole lot.  In general QTR does overkill on most printing matters to make it a non-issue.

So on Windows I think there potentially are small benefits to resampling to 720ppi before printing.  I can’t speak for Mac OS X, but I’d be more inclined to resample myself than to let the OS do it.

One way to avoid all these issues is to set the PPI to 720 and let the size adjust without resampling, and then print at that size.  Since there’s no resampling of the image then there can never be any resampling artefacts.  I’ve met people who do this.  Good luck to them.  I just can’t see that avoiding barely noticeable resampling artefacts is worth such a degree of inflexibility in print size.

This discussion has been about QTR and its use of 720ppi.  It’s my understanding that the Epson driver uses 360ppi, or 720ppi if the “finest detail” option is checked in the driver.  As far as I can tell, the same broad comments apply to it.  You can’t avoid resampling if you don’t send the image to print at the drivers native PPI, but whether there’s any visible benefit resampling yourself is not clear, although there may be on a Mac.

Much of what appears in this article is my summary of the discussion in this thread.

Last: Microbanding and Gloss  Next:  The Conversion Heretic

One thought on “Resizing and Resampling for Printing

  • 9 July 2021 at 2:45 am
    Permalink

    I tend to scan at the largest optical I can. Ive found that Jon’s recommendation to just let QTR handle the DPI when reducing an image size to work extremely well. My images are 300 dpi but often get reduced to 1000dpi or higher by QTR. I can’t see any difference between prints if I do this or send a reduced file at 300 or 720 dpi.

    Reply

Leave a Reply

Your email address will not be published. Required fields are marked *