Best Interpolation Process Qimage vs Photoshop - Poll

[Donald Cooper]

It is a nice phrase, but it does not apply always, not even close.
Qimage is probably not the best of them all, and certainly is not
magic, but it is up there. Jaggies are pretty easy to see, they are
not like a mild headache you think goes away because you put some
titanium/manget patch on your forehead. What Qimage does most
importantly, and people seem to miss this, is cut a lot of time,
offering comparable quality to a longer and much more carefully
done individual workflow.

Actually I agree with this completely.

As I said before, I recently had to print a list of more than a
hundred pics that I'm selling prints of. I was not gonna print a
whole set, so.. load qimage, drag the pics, set the size to 3"x2"
and print in letter sized paper that's it. No need to write an
action, resize or even arrange the photos in one letter sized
paper. Resampling quality is not even an issue here, but that it
saved so much time. I save files in png format instead of TIFF, and
you can still print in any color space you want (png does not carry
a color space tag), for every file because of the wonderful filters
which include color space override/assign. Countless of bits of
help that do not necessarily involve interpolation quality, which
is quite good too.

And this is the perfect sample. It's what I was trying to say when I mentioned that getting Qimage could be a good idea. You just said it better and with more detail.

Note that I'M NOT DISCUSSING QUALITY IN HIGH RESOLUTION PICTURES,
as this seems to be the usual misunderstanding. Just the fact that
interpolation is done at the driver. If you send higher res
pictures you probably won't notice the difference.

I agree with you on these points as well as your earlier point -- not quoted -- that the driver will interpolate. I was responding to people who think that they have to uprez an image from 359 ppi to 360 ppi before sending it to the printer.

 

[aruzinsky]

I have to admit those downsampled rings are really bad, but I see no point in showing that unless A is, in fact, "bicubic sharper." Did Ron say yet that A is "bicubic sharper?"

I see that. Entire ripple bands aren't missing, however, only a
tiny piece here and there and that might be caused more by the
antialiasing method on the downsample rather than the upsample.
Antialiasing, while it can sometimes remove a little detail that
could be called "real" detail, also helps to prevent a lot of
artifacts that create detail that shouldn't be there. Here is an
example of a downsample from 1000x1000 to 400x400 by both PhotoShop
CS2 bicubic sharper and Qimage pyramid with default settings. As
you can see, the PS version created a lot of "detail" that is
nothing more than aliasing artifacts. Of course, these are easily
controlled/removed in PS by first doing a gaussian blur of say 1
pixel, but PS's default downsampling doesn't include the
antialiasing. I think even for forensics, this is a tradeoff: is
the extra tip of a ripple worth the invented "detail" that is just
artifacts? Probably depends on what you are looking at/for, the
image itself, and other factors. In forensics, you normally want
all the detail you can get but on the other hand, you don't want to
see things that aren't really there either.



P.S. in the spirit of the original thread, BTW, the only reason I'm
not voting is that some may see my vote as biased.

--
Mike
Author: Qimage, Profile Prism
http://www.ddisoftware.com

--
Author of SAR Image Processor and anomic sociopath
http://www.general-cathexis.com

 

[The Davinator]

It's offered in Panorama Tools.

 

[Mike Chaney]

I have to admit those downsampled rings are really bad, but I see
no point in showing that unless A is, in fact, "bicubic sharper."
Did Ron say yet that A is "bicubic sharper?"

That I don't know. I guess he'll post that info once the voting is done (if anyone is really going to vote). :-/ I tried both bicubic sharper and bicubic smoother on the downsample though and didn't see much difference as far as the aliasing in the ring test is concerned.

--
Mike
Author: Qimage, Profile Prism
http://www.ddisoftware.com

 

[saynomore]

By the way, I didn't mean to imply that you were one of those people who "didn't get the point". It's just that there have been a lot of discussions (a couple of which I have been involved in) around Qimage's unsurpassed/lame interpolation quality, when to me that is not the most important feature of it.

have a nice day.

 

[saynomore]

try zooming out once and then out once again, it is cool. Sorry, it must be the medicine kicking in... It illustrates your point too though.

I see that. Entire ripple bands aren't missing, however, only a
tiny piece here and there and that might be caused more by the
antialiasing method on the downsample rather than the upsample.
Antialiasing, while it can sometimes remove a little detail that
could be called "real" detail, also helps to prevent a lot of
artifacts that create detail that shouldn't be there. Here is an
example of a downsample from 1000x1000 to 400x400 by both PhotoShop
CS2 bicubic sharper and Qimage pyramid with default settings. As
you can see, the PS version created a lot of "detail" that is
nothing more than aliasing artifacts. Of course, these are easily
controlled/removed in PS by first doing a gaussian blur of say 1
pixel, but PS's default downsampling doesn't include the
antialiasing. I think even for forensics, this is a tradeoff: is
the extra tip of a ripple worth the invented "detail" that is just
artifacts? Probably depends on what you are looking at/for, the
image itself, and other factors. In forensics, you normally want
all the detail you can get but on the other hand, you don't want to
see things that aren't really there either.



P.S. in the spirit of the original thread, BTW, the only reason I'm
not voting is that some may see my vote as biased.

--
Mike
Author: Qimage, Profile Prism
http://www.ddisoftware.com

 

[Mike Chaney]

If you zoom out 4 times and then zoom in 4 times, the image should be back to 100%. Well, it is, but it gets all chunked up and the tiles don't align properly. Looks like Phil might have a bug in his zooming code.

--
Mike
Author: Qimage, Profile Prism
http://www.ddisoftware.com

 

[saynomore]

If you zoom out 4 times and then zoom in 4 times, the image should be back to 100%. Well, it is, but it gets all chunked up and the tiles don't align properly. Looks like Phil might have a bug in his zooming code.

--
Mike
Author: Qimage, Profile Prism
http://www.ddisoftware.com

 

[Ron AKA]

One of the suggestions made in this poll, was that it was unfair to compare these interpolation methods on screen, and the print was the real test. So to test that theory I did prints of the actual images in the initial post, scanned all of them at 720 ppi, resized using Photoshop Bicubic Sharper to 100 ppi and 4.8" wide for posting here. All prints were done at 1200 ppi maximum. Prior to printing they were treated differently to reflect the two processes:

PhotoShop - Prints were resized to 300 ppi and 4" wide with the same method used in the original up sample. Then they were sharpened using PhotoShop Elements 4.0 Auto setting and printed using PhotoShop.

Qimage - Resolution was left alone and the prints were made to 4" using the orignal up sample method to the printer native ppi (1200), and autosharpening on at the default level.

As an aside I downloaded a trial copy of Elements 4.0 to do this (all 350 MB!), and while the print interface has improved from 1.0, it is still miserable compared to Qimage. Multiple prints per page are now somewhat possible, but I finally gave up in laying them out as I wanted, and decided paper was cheaper than my time and printed one print per page.

My conclusion? I think this essentially proved nothing, other than when one prints and scans some quality is lost. The prints in my hand look essentially like the images I see on screen in the original post The hard copy is easier to compare, and I will post my semi-blind preferences of the prints and the screen versions later. Some are so close that I'm sure the order will change a bit.

Original (not printed or scanned, and this is the standard to which the scans should be compared)



Method A



Method B



Method C



Method D



Method E



Method F



Ron

 

[Bill Waterson]

F looks the best.
--
Bill

 

[The Davinator]

All you've done is magnify them back for the screen....thus getting the same result as the on screen original. As I said, what is on screen in 2.5X or more larger than what is on print. The artifacts you keep showing with F vs A vanish.....leaving A looking sharper.

I can see that this is a pointless task when people insist on comparing apples to crowbars.

 

[Mike Chaney]

I think what this proves is that if you print them so that they are the same physical size as they appear on screen, you'll probably prefer the same method both in print and on screen. The only time you might choose something different in print is when you shrink the result down and create a smaller print. No surprise there as shoving the same number of pixels into a smaller area changes the balance of artifacts to sharpness and therefore the entire interpolation dynamic.

--
Mike
Author: Qimage, Profile Prism
http://www.ddisoftware.com

 

[The Davinator]

True Mike.

My point is that when interpolated to say, 360 dpi for print.....viewed at 100% on screen, your 20x30 actually becomes the equivalent of at least a 113" print. We're not talking about "shrinking" the image....we are talking about viewing it on screen the same size as one does on print. This print example just above, doesn't do that. It in fact makes the image on screen even bigger.

This gets into the pixel peeping issue that I find ridiculous. If interpolated to 20x30 & 360dpi, the image on screen would have to be reduced by a factor of 3.75. This makes all the artifacts people rave about vanish. I see that you understand this....but I'm finding very few others seems to grasp this very simple fact of life for those of us outputting large prints.

 

[Ron AKA]

I don't think you get the purpose of the test. Let me review.

We start with a pretty clean but not real high resolution image, but one I think most would find quite pleasing to the eye - both on screen and on paper at about 4" wide.

Next we destroy about 90% of the data (118k pixels down to 1.2k) to get a very low resolution image. The whole purpose of the test is to evaluate a few different interpolation methods to see how well each can recreate the original image. The closer the results look compared to the original the better. That is the only question that needs to be asked - how close does it look? If details are lost or artifacts are added then that is obviously a negative effect. No method of course is going to be perfect, because you are making data where there is none.

Some seem to have focused on viewing and printing the low resolution print or interpolated print at a small size. This is meaningless. The whole purpose is to recreate an image from the limited data set that is the same size as the original and looks the as close to the same as the original as possible. That is what interpolation is used for. It is the cover up for not really having enough resolution for a good print.

To me it is pointless to jsut up sample an image and only compare the upsampled images to each other. You have no original to be the absolute measure or control point, so it all becomes subjective. So I think this is a totally fair test, and as shown in the last step printing is not the issue. This is without a doubt an apples to apples comparison of interpolation methods. Yes, you could say you need a different method for each different original image. Probably true that some may favour certain shapes , but in my experience the same ones keep coming to the top and the same ones to the bottom even when different images are used.

Your point seems to only be that if you make the image small enough quality does not count for much. I would call that a true blind test. If you can't see it, then any comparison is just subjective. No interest to me.

Ron

 

[Ron AKA]

Dave, yep you don't get it. How do you explain why the good ones look so good? They don't seem to be pixel handicapped!

 

[Mike Chaney]

There is one thing that must be pointed out when you talk about 113" prints, 20x30 prints, and on screen sizes versus print sizes. Both display and print size are completely arbitrary. None of us are forced to decide what size we are going to print before we take a photo. That is something that should be decided upon and dealt with later. When quoting sizes, you often assume that we are printing and reviewing the entire image. This is OK if the splash photo in this thread is the entire frame. Don't forget that we often crop parts of an image, and we often focus on small portions of an image when viewing them. "Good photographers" take shots so that only minimal cropping is necessary of course, but occasionally we find an unexpected artistic quality in a small portion of the image and choose to do an extreme crop. These are just facts of [photographic] life. Just as an example, the water drop shown in this thread could be part of a larger image: one showing a park with people standing around a water fountain and this is just a zoom-in of one water drop at the edge of the fountain. You don't necessarily have to crop the water drop and print it: it could simply be detail that you want to see in a larger blowup of the whole image. If you print a 20x30 print from a 6 MP camera like a Canon 10D, you only have 100 PPI to start with so a 6x or more upsample to the printer PPI is perfectly legit. Do a little cropping, and you'll need more than that.

My point is that there is a size where acutance of a print is roughly equivalent to what you see on screen. Though I don't necessarily agree with your 3.75x because this is further complicated by display technology (continuous tone versus dithered), viewing distance, vision and other factors, I see that you agree with the overall concept. What I take issue with are your quoted sizes like 113" and 20x30. Those are completely arbitrary because we could be looking at any (smaller) portion of an image, any screen size, and any print size. And since there is a print size that correlates nicely to any displayed size on screen, I see nothing wrong with comparing interpolation algorithms on screen. In going back to what I said earlier, there is a print size that correlates nicely to what you view on screen. Further, given the range of sizes, viewing distances, amount of area in the image covered by the detail in question, and other factors, comparisons like these are not at all ridiculous like your 113" quote might suggest. Sure, while taking an on screen image and squashing it down into a tiny print might obscure artifacts, printing it even larger will show them more. It's all relative.

The upsampled choices in the original thread (A-F) are 4.25 inches across on my screen. Personally, for me, if I print the samples at the same size (4.25 inches across), I see the same artifacts/sharpness in print as I see on screen. To me, that's a 1:1 correlation regardless of the fact that the print contains more physical printer "dots" as defined by the printer than the screen contains in the same area. Of course, if you take the same image and dump it to the printer so that one pixel in the image correlates to one printer dot on the page, you'll end up with a shrunken tiny print due to the higher pixel density of the printer, but I don't think that's what we should be comparing here.

--
Mike
Author: Qimage, Profile Prism
http://www.ddisoftware.com

 

[The Davinator]

"Your point seems to only be that if you make the image small enough quality does not count for much. I would call that a true blind test. If you can't see it, then any comparison is just subjective. No interest to me."

Ron,

You are missing the point. Comparing full rez screen shots vs print are NOT the same. If you interpolation test is to do nothing more than view on screen, then you'd be correct......although your test would then be a waste of time. I'm not talking about small prints. You continue to miss the obvious relationship between screen vs print.

Mike seems to disagree with me as well. However, the math I used to reach print size is pretty easy to grasp. Screen rez of 96 vs interpolated print size of 360. 360/96 = 3.75X. There is no voodoo magic here. Now while a few here have been playing with screen shots of spiral shapes, etc, I ran a few prints of real photos (both digital and scanned 4x5) and cropped out 4x6 shots. Printed on Canon high gloss. To the naked eye, there is NO DIFFERENCE between a 360 dpi image from PS and Qimage sending a 600 dpi or 720 dpi (depending on printer) file size to the printer. I've played with this with enough people to have found NO ONE could see a difference.

I've mentioned before having a test run at both 16x24 and 20x30. We have found Qimage Pyramid to be a bit softer. How much? You'd have to almost have your nose touching the print to see it. And the artifacts you're going on about? Invisible.....even with the nose nearly touching the print at both 16x24 & 20x30. Did the artifacts show on screen? Yes. But as I said, at 100% on screen, you've overmagnified the image when compared to print.

I'm surprised some people have such a great deal of difficulty with this. It's easy.....and is mentioned in printing workshops by people like Charles Kramer and George Dewolfe. Upsample an image to 360 dpi. Print a crop. View the crop print against what you see on screen at 100%. You'll quickly see you need to reduce screen size WAY DOWN to match print and screen size.

So please, Ron. Stop telling me how I'm talking about small prints. I'm not. I was talking about a 20x30.....pretty large by most people's standards. Stop viewing at 100% on screen, and stop showing prints scanned in a matter that they would have to be viewed under an 8X loupe to see the artifacts you're talking about.

I'm done with this. This is just another perfect example of forum experts who ramble on about math , etc, rather than using their eyes. I suggest you attend a fine art printing workshop by Mr. Kramer or Mr. Dewolfe. You'll find out how silly this on screen 100% viewing contest really is!
Best wishes for everyone in the new year.

 

[Mike Chaney]

Mike seems to disagree with me as well. However, the math I used
to reach print size is pretty easy to grasp. Screen rez of 96 vs
interpolated print size of 360. 360/96 = 3.75X. There is no
voodoo magic here.

What size is your screen? What resolution are you running on your screen? Mine is 15 inches wide and I run it at 1600x1200 which equates to 107 PPI. If I decide to run it at 1024x768, the screen doesn't change size (still 15 inches wide) so I'll be at 68 PPI. Take another screen that has a smaller footprint with a width of 13 inches and run it at 1600x1200 and you'll be at 123 PPI. Now, how close do you sit to your screen? How did you display the image on screen? Did you display it full screen? In a window? How big did you print it and how close do you hold the print to your eyes? Is your vision the same when holding a 4x6 print 9 inches from your eyes as it is when viewing your 21 inch monitor at 2 feet distance? How does the continuous tone on the monitor differ from the halftoning on the print and how does that affect perceived clarity?

Surely you see how arbitrary this is and that you cannot just pull a number out of your butt like 3.75! Yes you can get more pixels in the same physical size with a printer and yes you may have to display them at a different size than you print them to get the same overall appearance. I don't think that was ever the point of this thread. Sizing is arbitrary and should be arbitrary. There are many factors involved, not just PPI of the screen and PPI of the print driver. Your argument about printers and screens being run at different PPI is only valid if we force ourselves to dump an image to the printer at 360 PPI (for example) and don't allow ourselves to pick a size or "stretch" the pixels. That would be like saying you can only get one print size out of a 10D: 8.33 x 5.56. Mapping pixels in the image to pixels on the output device at a 1:1 ratio is not a requirement of any kind of photography I'm aware of.

--
Mike
Author: Qimage, Profile Prism
http://www.ddisoftware.com

 

[The Davinator]

Mike seems to disagree with me as well. However, the math I used
to reach print size is pretty easy to grasp. Screen rez of 96 vs
interpolated print size of 360. 360/96 = 3.75X. There is no
voodoo magic here.

What size is your screen? What resolution are you running on your
screen? Mine is 15 inches wide and I run it at 1600x1200 which
equates to 107 PPI. If I decide to run it at 1024x768, the screen
doesn't change size (still 15 inches wide) so I'll be at 68 PPI.
Take another screen that has a smaller footprint with a width of 13
inches and run it at 1600x1200 and you'll be at 123 PPI. Now, how
close do you sit to your screen? How did you display the image on
screen? Did you display it full screen? In a window? How big did
you print it and how close do you hold the print to your eyes? Is
your vision the same when holding a 4x6 print 9 inches from your
eyes as it is when viewing your 21 inch monitor at 2 feet distance?
How does the continuous tone on the monitor differ from the
halftoning on the print and how does that affect perceived clarity?

This is actually all not relevant to what I'm saying. What I'm saying is that you can't compare an image interpolated to 360dpi at 100% on screen, and think that is what your print will look like.

Surely you see how arbitrary this is and that you cannot just pull
a number out of your butt like 3.75!

How ignorant are you? My screen rez works out to 96 ppi. 360/96 = 3.75X

I really don't care what your screen rez is. In your case it would be 101" and thus a factor of 3.37......so you see, you still need to reduce it a large amount. Who knows what point you're trying to make.

Yes you can get more pixels

in the same physical size with a printer and yes you may have to
display them at a different size than you print them to get the
same overall appearance. I don't think that was ever the point of
this thread. Sizing is arbitrary and should be arbitrary. There
are many factors involved, not just PPI of the screen and PPI of
the print driver. Your argument about printers and screens being
run at different PPI is only valid if we force ourselves to dump an
image to the printer at 360 PPI (for example) and don't allow
ourselves to pick a size or "stretch" the pixels. That would be
like saying you can only get one print size out of a 10D: 8.33 x
5.56. Mapping pixels in the image to pixels on the output device
at a 1:1 ratio is not a requirement of any kind of photography I'm
aware of.

Well, now you're aware of it. At 360dpi, you can only get one print size from a 10D.....of the measurements you gave......unless you interpolate. Now when you do, the ratio between screen & print changes by the ratio of the interpolation.....the range gets larger as you increase the print rez output size from 300 to 360 to 600 to 720, etc when viewing on screen at 100%

I'm truly surprised you don't understand such a basic concept of printing.

I'm sorry Mike, but if you don't understand such a basic concept, there is no point is discussing anything further with you. I suggest you do some reading up on the topic....or better yet, do as I said and make a print and compare it to your screen @ 100%. Rather than telling me I'm wrong, just do it. Then maybe you won't come across such a fool to those of us who have tried it and understand this simple concept. Or as I said to Ron....attend a couple of workshops from other people who have figured this out. I mentioned George Dewolfe and Charles Cramer as two of the most experienced high quality printers out there (I misspelled his name "kramer" earlier.) Once you do, I'm sure you'll find, like I did, that most of what you'll see in the Printing forum comes from uninformed inexperienced forum experts. I had hoped you wouldn't fit into that category.....but now I'm beginning to wonder.

--
Mike
Author: Qimage, Profile Prism
http://www.ddisoftware.com

 

[Mike Chaney]

That'd Dave! Always taking the discussion to a lower plane. I'll leave you here to talk to yourself now. You seem to love to hear yourself talk and apparently everyone on this forum gets it but you anyway. I see no need to keep trying to educate you when you are already a self proclaimed expert. I take the time to explain all the variables involved and you continue to use your third grade math to try to compare things that are not comparable. 3 apples divided by one orange equals a potroast. Yeah, okay Dave. Ramble on.

--
Mike
Author: Qimage, Profile Prism
http://www.ddisoftware.com