PPI > DPI total confusion

[digitaldog]

Page 130 of Jeff Schewe's "The Digital Print".

Or:
https://www.digitalphotopro.com/technique/photography-workflow/the-right-resolution/


The people at Epson say that the print driver doesn’t do the re-sampling, and since the application sending the print doesn’t do the resampling unless asked to do so, the resampling must be happening in the print pipeline. I’ve tried, unsuccessfully, to get confirmation from Apple and Microsoft about their respective print engine activities regarding the resampling of the image data. So I really don’t know where or how the resampling is being done. But I’m convinced some sort of resampling is being done. Is it an optimal resampling algorithm, or is it something done for speed? I don’t know, but I suspect, at best, it’s a compromise in favor of speed. I’m pretty sure there are better, optimized resampling algorithms that could do a superior job. In fact, Adobe Photoshop Lightroom resampling is a hybrid Bicubic algorithm that interpolates between Bicubic and Bicubic Smoother for upsampling and Bicubic and Bicubic Sharper for downsampling.

[Jim Kasson]

I'm not sure test prints really encapsulate the dimensions of the LUT in play here.  At some point you obviously need to test, but I don't believe there's a suitable series of prints to be used to create LUTs of this complexity.

When I make LUTs, I used several thousand color patches and  read them with an iSis XL

Jim

[Jim Kasson]

The examples on your blog are excellent, but also simple.  You're not looking at multiple colour variations with an expectation of consistent rendering on a reflective (rather than transmissive) media across multiple substrates, inks, and lighting conditions.

Every time you change halftoning algorithm, inkset, media, or -- to some extent -- print illumination, you need to make new LUTs.

Jim

[Jim Kasson]

Resampling in other than linear space can create varying levels of moire as it introduces harmonics which reflect downward of off Nyquist. Ideally, resampling upward should not alter the spectral components of images. There are resampling techniques that try to fill in higher frequency information based on assumptions about what natural images contain. There be dragons.

Caveat. Most of what I have done is, for instance, creating specialized charts of different sizes that will produce the same spectral characteristics when photographed at an appropriately scale distance. Images that are output referred are already munged (non linear intrinsically) and linear space up-sampling them is a different beast unless one can somehow revert them to scene referred, upsample, then convert back to output referred. Yuck. At that point it's really more a question of artistic objective.

Some time ago, I did a lot of work lots and lots of help from Bart van der Wolf and Nicholas Robidoux on downsampling. I can provide references and we can talk about the proper nonlinearity for the computations if you wish, but that would be series thread drift.

But, as I said earlier, if you use NN, as the Epson driver does, it doesn't make any difference. Resample a PPRGB image directly or convert it to a linear space with the same primaries and WP, resample, and convert it back and you're going to get the same answer.

Jim

[Jim Kasson]

Or:
https://www.digitalphotopro.com/technique/photography-workflow/the-right-resolution/


The people at Epson say that the print driver doesn’t do the re-sampling, and since the application sending the print doesn’t do the resampling unless asked to do so, the resampling must be happening in the print pipeline. I’ve tried, unsuccessfully, to get confirmation from Apple and Microsoft about their respective print engine activities regarding the resampling of the image data. So I really don’t know where or how the resampling is being done. But I’m convinced some sort of resampling is being done. Is it an optimal resampling algorithm, or is it something done for speed? I don’t know, but I suspect, at best, it’s a compromise in favor of speed. I’m pretty sure there are better, optimized resampling algorithms that could do a superior job. In fact, Adobe Photoshop Lightroom resampling is a hybrid Bicubic algorithm that interpolates between Bicubic and Bicubic Smoother for upsampling and Bicubic and Bicubic Sharper for downsampling.

Here is an old test I did that proves that the Epson resampling is NN:

http://blog.kasson.com/technical/injet-printing-on-epson-part-3/

Jim

[Doug Gray]

Some time ago, I did a lot of work lots and lots of help from Bart van der Wolf and Nicholas Robidoux on downsampling. I can provide references and we can talk about the proper nonlinearity for the computations if you wish, but that would be series thread drift.

I'm curious about this but, yeah, it's thread drift.  I'll PM.

But, as I said earlier, if you use NN, as the Epson driver does, it doesn't make any difference. Resample a PPRGB image directly or convert it to a linear space with the same primaries and WP, resample, and convert it back and you're going to get the same answer.

No disagreement. I didn't mean to suggest that in regards to printer NN interpolation.

Jim

[Doug Gray]

Here is an old test I did that proves that the Epson resampling is NN:

http://blog.kasson.com/technical/injet-printing-on-epson-part-3/

Jim

I just checked the Canon. It's also NN.

[Jim Kasson]

I just checked the Canon. It's also NN.

Thanks, Doug. Good to know.

Jim

[Rand47]

Or:
https://www.digitalphotopro.com/technique/photography-workflow/the-right-resolution/


The people at Epson say that the print driver doesn’t do the re-sampling, and since the application sending the print doesn’t do the resampling unless asked to do so, the resampling must be happening in the print pipeline. I’ve tried, unsuccessfully, to get confirmation from Apple and Microsoft about their respective print engine activities regarding the resampling of the image data. So I really don’t know where or how the resampling is being done. But I’m convinced some sort of resampling is being done. Is it an optimal resampling algorithm, or is it something done for speed? I don’t know, but I suspect, at best, it’s a compromise in favor of speed. I’m pretty sure there are better, optimized resampling algorithms that could do a superior job. In fact, Adobe Photoshop Lightroom resampling is a hybrid Bicubic algorithm that interpolates between Bicubic and Bicubic Smoother for upsampling and Bicubic and Bicubic Sharper for downsampling.

Andrew,

GREAT to see you post here!!!

Rand

[rdonson]

Andrew,

GREAT to see you post here!!!

Rand

Ditto!!!

[Ernst Dinkla]

Thanks, Doug. Good to know.

Jim

A decade ago I used the Qimage Ultimate resolution targets to check whether the HP Z3100 driver etc had a decent anti-aliasing routine on downsampling. It did not, aliasing was very visual, NN resampling most likely. An old message on that subject;

https://groups.yahoo.com/neo/groups/QuadtoneRIP/conversations/messages/7073

Met vriendelijke groet, Ernst

http://www.pigment-print.com/spectralplots/spectrumviz_1.htm
March 2017 update, 750+ inkjet media white spectral plots

[digitaldog]

Here is an old test I did that proves that the Epson resampling is NN:

http://blog.kasson.com/technical/injet-printing-on-epson-part-3/

Jim

Sorry, I'm not clear on the meaning of NN.

[Jim Kasson]

Sorry, I'm not clear on the meaning of NN.

Nearest neighbor

[digitaldog]

Nearest neighbor

Ah thanks, I should have figured that out. 

So this agrees with Schewe's text and the idea one should resample using something much better then send the data to the print driver.

But what's producing this NN? Jeff states it's the OS pipeline if we are to believe what Epson wrote (it's not the driver).

[digitaldog]

Oh, and like others, I don't see the 340PPI example Jim.

[Mark D Segal]

Ah thanks, I should have figured that out. 

So this agrees with Schewe's text and the idea one should resample using something much better then send the data to the print driver.

But what's producing this NN? Jeff states it's the OS pipeline if we are to believe what Epson wrote (it's not the driver).

This isn't a new discussion and usually the point at which it gets stuck for lack of vendor transparency. What's the big secret? There are only four places where resampling can happen: the application, the CMM in the OS, the printer driver, or the printer firmware. Between Adobe/Apple/MS/Epson why can't they just come out with a coherent story-line on how this works and where the operation is performed? I'm not sure what it buys us to know for sure, but at least to satisfy curiosity about the tools we are using, and perhaps helpful for some diagnostics here and there, who knows.

[Schewe]

Between Adobe/Apple/MS/Epson why can't they just come out with a coherent story-line on how this works and where the operation is performed?

Because I'm pretty sure no one person actually knows the answer because each component; the application, the print driver and the OS are all being done by groups of people who are only responsible for subcomponents and don't know or understand the entire pipeline. Also not that the Mac and Windows pipelines are different and I think the drivers for each OS are different.

But, for the purposes of our discussion, it's 360/300 PPI for Epson/Canon or HP and 720/600 PPI for certain driver configs. That's pretty much proven, right :~)

[Mark D Segal]

Because I'm pretty sure no one person actually knows the answer because each component; the application, the print driver and the OS are all being done by groups of people who are only responsible for subcomponents and don't know or understand the entire pipeline. Also not that the Mac and Windows pipelines are different and I think the drivers for each OS are different.

But, for the purposes of our discussion, it's 360/300 PPI for Epson/Canon or HP and 720/600 PPI for certain driver configs. That's pretty much proven, right :~)

Jeff, there must be people who have a good overall command of what happens where. This has to be essential for each company to know what it needs to design and how to design it.

I agree the answer to the original concern is what you say above.

[Farmer]

But, for the purposes of our discussion, it's 360/300 PPI for Epson/Canon or HP and 720/600 PPI for certain driver configs. That's pretty much proven, right :~)

It's 360/720 or 300/600 (P20000/10000) for Epson.

My understanding is that in OS X, the OS renders and uses a unique interpolation algorithm similar to bi-cubic or bi-linear (it's done via Quartz Compositor).  In Windows, it's done by the driver and is based on NN.

Also, in the Settings dialogue, don't choose "Processed by Printer" - that's not to do with resolution directly, but it makes a decision about whether to send RGB to the printer for conversion to ESC/P or whether to do it on the computer within the driver.  Sending it to the printer means the RGB is sent as a JPG (i.e. lossy compression).  So don't do that :-)

[Jim Kasson]

Oh, and like others, I don't see the 340PPI example Jim.

Andrew, I'm sorry about that. There was a WordPress upgrade that rendered some in-line images unviewable. I really should go back and re-do the test.

Jim