How Sharp Is Your Printer? How Sharp Are Your Eyes?

[bjanes]

In general you cannot optimize a single print detail-wise for all viewing distances. If you prepare an image so it looks perfect at a few inches away (sharp, natural, no artifacts) then it might look fine much farther away, but it won't be as sharp as it could be. That is, you could sharpen the image a lot more if you knew you'd be viewing it from no closer than 10 feet away. But then that same image would look bad when viewed at 10 inches.

Eric (and Bart),

For discussion, let's assume that we want to sharpen for a viewing distance of 34 cm (13.5 inches), to coincide with the conditions given by Bob Atkins in his discussion of SQF. The eye can resolve down to 30 cycles/degree (60 lines or 30 line pairs per degree), but is most sensitive as far as apparent sharpness is concerned to about 6 cycles/degree. At a viewing distance of 34 cm, these frequencies correspond to 1 cycle/mm and 5 cy/mm respectively. 5 cy/mm is 127 cy/inch or 254 lines/inch and is in agreement with Jeff Schewe's figures.

To sharpen for the optimal MTF of the eye at this viewing distance would seem to require parameters that optimize the image at 1 cy/mm. How do we accomplish this? Jeff would set the sharpening radius to to keep the sharpening halos to about 0.01 inch, but does not take the viewing distance into account. How does this correspond to the supposed ideal sharpening for this image?

It is interesting to note that for a full frame 35mm camera, the image on the sensor must be enlarged 8 times to get an 8 x 12 inch print, so  the required resolution in the plane of the sensor to obtain the above resolutions would be 8 cy/mm and 40 cy/mm respectively and that these resolutions correspond to the 10 cy/mm and 40 cy/mm often given for MTF for 35 mm lens testing.

[artobest]

Am I the only one who is failing to completely understand this thread?

[Ernst Dinkla]

Am I the only one who is failing to completely understand this thread?

No, I do not grasp all of it either.
But based on what I have seen of Bart's messages in the past, I very much trust his knowledge.

Then there is that aspect of how to translate the theory reasonably efficient to daily practice.
Using Qimage here with Z3100 and Z3200 printers and always making proof strips for the larger sizes before the actual print is made.
With the knowledge that it seldom happens that there is plenty of image data at that size.


met vriendelijke groeten, Ernst Dinkla

Try: http://groups.yahoo.com/group/Wide_Inkjet_Printers/

[Bart_van_der_Wolf]

In general you cannot optimize a single print detail-wise for all viewing distances. If you prepare an image so it looks perfect at a few inches away (sharp, natural, no artifacts) then it might look fine much farther away, but it won't be as sharp as it could be. That is, you could sharpen the image a lot more if you knew you'd be viewing it from no closer than 10 feet away. But then that same image would look bad when viewed at 10 inches.

Hi Eric,

Obviously one cannot optimize for all conceivable viewing distances at the same time. Obviously one would like to optimize for a given viewing distance, because it is relatively easy to do (once one considers the Contrast Sensitivity limitations of human vision).

Does that mean that we cannot do anything, even with our knowledge of the CSF? IMHO the easy answer is not the correct answer, and I hope to demonstrate it with the following (sharpened) adaptation of the image I posted before:


Same test as earlier/above, but when you now increase the viewing distance to the screen, you'll see that the low contrast high spatial frequency detail stays visible much longer, arguably even at almost the same level as larger detail, untill one exceeds the critical viewing distance far enough. The explanation is that the image has received a "multi-resolution" sharpening. The MR-sharpening effectively consisted of an increasing level of sharpening as the spatial frequencies got higher.

That resulted in an increased amplitude of the higher frequencies as illustrated by this horizontal cross section plot of amplitude:


Is the sharpening perfect? No, there are artifacts visible. Part of the problem lies in the fact that the original detail was already partially aliased, and that was only brought out better by sharpening. BTW that is also why it important to print at the native resolution of one's printer (driver) usually 600 or 720 PPI, and sharpen at that final PPI level. The Shannon/Nyquist theorem states that we require more than 2 lines to reliably samle a single line.  Sharpening in general will make aliasing visible, but at these high PPI levels the aliases will be less present/visible. I know that some participants of this forum are of an opinion that there is no benefit to be expected from upsampling to the native PPI (which can vary with paper choice) of the printer, but I obviously don't agree.

Another part of the problem is that we are looking at a computer screen with very poor (usually 96 PPI) resolution, so even the finest detail is very easily resolved by our eyes. The target itself of course also presents a worse case torture test, and I exaggerated the amount of sharpening to drive the concept home, and we can also expect larger losses due to the print process than by viewing a display. Doing it in a more subtle way could be used for web publishing, but then we also have to consider a usually predictable close viewing distance, which reduces the need for a multi-resolution approach.

I hope this demonstration will open the eyes of some of those who tend to debate from an entrenched position. The glass is half full, there are more possibilities than are currently being used.

Cheers,
Bart

[Ray]

Obviously one cannot optimize for all conceivable viewing distances at the same time. Obviously one would like to optimize for a given viewing distance, because it is relatively easy to do (once one considers the Contrast Sensitivity limitations of human vision).

The fact is, photographers tend to want their prints to look good from any distance. That's presumably why some photographers spend tens of thousands of dollars on a 60mp MFDB.

The photograph is different from a painting in respect that the large painting really needs to be viewed from a good distance to be apreciated, especially impressionistic paintings.

Any photographic print, however large, will invite the photographer to inspect it from a close distance to see the resolution.

In the real world, the closer you get to any subject, the more detail you see. If your eyesight's not good enough at close distances, wear spectacles and you'll see more detail. Peer through a microscope and you'll see even more detail.

The photograph attempts to capture reality, or at least the reality we saw from the position we were when we took the shot. There's a certain fascination in being able to see more detail the closer you get, just as in the real world.

In fact, in the real world you may never get close enough to see the veins in the eyeball of your favourite movie star. A photograph may capture such detail so you can get close enough.

An impressionistic painting from close up, tends to look like crap, just a jumble of brush strokes. Likewise, a large advertising poster from across the highway might look terrific from a distance of 30 metres. Inspect it close up, and all you see are colored dots the size of small peas.

Perhaps the significant question here is, having optimised a print for viewing at a large distance, is it also possible to make it presentable at a close viewing distance without compromising those qualities appreciated at the larger distance.

Bart implies that it is, but I'd like to see some real-world photos demonstrating this. My feeling is that the exaggerated contrast of the print that makes it suitable for distance viewing, will tend to look like crap at close distances.

[howseth]

I agree with Ray: there is a fascination in photography of so much sharp detail that the closer you get the more you can see - that is a difference with most painting (or TV and video). Balancing out all the wonderful detail with the question of contrast and the overall general beauty of a print from various viewing distances can not be reduced to a single formula for everyone - each photographer has to figure what works for their particular imagery and taste. Just like in selecting which type of paper to use, which type of camera system - and for some (like me) ultimate sharpness is not even desired. (though I did generally work with large format view cameras for much of my work)

Howard

[Deepsouth]

The Multi-Resolution Sharpening Bart mentions, is this the same as the Adaptive Unsharp plug-in? I did a brief web search on Adaptive Unsharp and it sounds like they might be the same.

[Bart_van_der_Wolf]

The Multi-Resolution Sharpening Bart mentions, is this the same as the Adaptive Unsharp plug-in? I did a brief web search on Adaptive Unsharp and it sounds like they might be the same.

Hi Deepsouth,

I don't know what that plugin does, but I assume it doesn't sharpen smooth areas such as the sky, a sort of built-in masking.

The functionality I'm mentioning targets multiple resolution levels in a given image at the same time. I'm working on a special approach, but for the example I used a special sharpening function of a program for astronomical image processing called ImagesPlus in a novel (viewing distance compensating) way. Maybe I should patent it ...

When I can find a little time, I'll prepare an image crop that has been processed this way and make it available for either display or print purposes. That way people can try it themselves on their printer. One of the issues is that different printers deteriorate the image in different ways, so I'll have to figure out something that on average still works okay.

Cheers,
Bart

[Ray]

When I can find a little time, I'll prepare an image crop that has been processed this way and make it available for either display or print purposes. That way people can try it themselves on their printer. One of the issues is that different printers deteriorate the image in different ways, so I'll have to figure out something that on average still works okay.

Cheers,
Bart

Bart,
I wonder how significant this issue is. If anyone wants an image to look good from both afar and close up, then the first requirement is to get a camera with a high pixel count. In fact, the P65+ is the current ideal candidate.

With such a DB, I imagine one could produce a 36" x 48" print which looks as great as a print from a humble 12.7mp 5D, viewed from a distance of, say 3 metres. But, on close inspection the P65 print will exhibit far more detail.

If one considers the print as a work of art, like a painting, the fact that it also looks good at a closer distance than one would view the print to appreciate the work as a composition, is a sort of bonus, but perhaps not the main point.

Nevertheless, if one can tinker with sharpening routines so that the result from both afar and close up is not compromised, then that's ideal.

[Ernst Dinkla]

The functionality I'm mentioning targets multiple resolution levels in a given image at the same time. I'm working on a special approach, but for the example I used a special sharpening function of a program for astronomical image processing called ImagesPlus in a novel (viewing distance compensating) way. Maybe I should patent it ...


Cheers,
Bart

Bart,

Like your comments on anti-aliasing routines improved Qimage's downsampling some years ago this could improve Qimage on smart print sharpening features. I wouldn't mind if that means a Qimage "Extreme" Studio version for an extra 10$. It is possible that the big money is at the other side, several  representatives of that side on this forum :-)


met vriendelijke groeten, Ernst Dinkla

Try: http://groups.yahoo.com/group/Wide_Inkjet_Printers/