Best Interpolation Process Qimage vs Photoshop - Poll

[Ron AKA]

With the one hot thread of late, perhaps there is no appetite for this, but as an engineer the subject facinates me. It is a formidable challenge to take a limited sample of data and try to recreate the original image. Purists were highly critical of digital music when it first came out ... and perhaps some still are, but one cannot argue that with correct sampling, Dolby and the like that there are not some pretty spectacular results. Digital sampling to recreate an analog output is the game in audio, and now in digital photography.

So if you are up to it, the following images have been downrezed from the original 5.5" width and 72 ppi, to 1.75" at 72 ppi, and then uprezed back to the original size and at 72 dpi. The interpolation method used for both the downrez and the up rez were the same. In the case of the Qimage process, the downrezed image was saved in TIF and then convered to JPG at the final step. For Photoshop the image was not saved until the final uprez step had been compled. So both methods only resulted in one JPG save.

The first image is the original, which was chosen to include gradual curves, and soft transitions to test the various methods - a drop of water splashing and creating ripples. If you wish to post your preferences in order, and there are sufficient responses it may be interesting to see what the poll concludes. There are a total of six methods; three from Qimage and three from Photoshop (including Bicubic Smoother). The order is random by draw. One can see some obvious differences expecially from the original in sharpness and contrast. No corrections were made to either in the process.

Here's hoping I get actual images to show this time!

Original:



Image A



Image B



Image C



Image D



Image E



Image F

 

[The Davinator]

Ron,

Part of the problem with posts like this is that, as with the last thread on this same issue, a lot of people (Palmswestphoto & Baladev) don't understand the relationship between what they see on screen @ 100% and what they see on print.

An image rezzed to 240 dpi would on print,would need to be reduced by about 60% on screen to look the same. So while we have people harping on about supposed haloing, high local contrast, and overly aggressive accutance artifacts, they don't undertand that when viewed on screen, these need to be reduced to compare what it would look like in print.....even more so if the image was rezzed to 300 or 360 dpi.

When viewed on print at sizes like 16x24.....which by the way is probably bigger than 99% of the people on these threads print......these apparent artifacts vanish and leave us with a sharper looking print.

Properly executed and sharpened images using Pyramid, Sync 256, S-Spline, Bicubic smoother , etc, etc. etc, will have only the most minor differences on print. I give the nod to properly executed bicubic routines along with S-Spline & sync 256 over lanczos, vector, and pyramid. But as I said, these differences at even 16x24 are extremely small. If you print smaller than that, you can get away with pretty much anything.

But until the testers understand the differences in spacial relationships between what they see on screen @ 100%, and what they get on print, we're going to end up with another load of garbage for this thread. In other words, a bunch of people who have either never done large scale printing, or those who never looked at the screen vs their print, will descend on the thread telling those of us actually comparing correctly, that we're wrong.

Let's see how this one turns out!

 

[fotografer]

I am assuming we are looking at the interpolations for their ability to produce 'natural' transitions? I can't say for sure how they look like when they are printed, but viewing purely ON-SCREEN , I prefer:

1) Image F
2) Image C
3) Image E
4) Image D (followed closely by)
5) Image B
6) Image A

--
fotografer

 

[fotografer]

So, what are your choices, Dave? C'mon, lighten up, let's have some fun (I think you like the high-contrast kinds, but you can qualify and say these come up better when printed, whatever, just CHOOSE!).

--
fotografer

 

[The Davinator]

Fotografer,

Actually, it's a tense moment here.....Magenta clog going on the proof printer (canon 9900) given me a headache!

 

[fotografer]

Oh dear, Dave, clogging at this time of the year?!

Anyway, I may post a difference set of preference again once I get them all to print with my lowly HP ps7150 (yep, almost three years and still going strong)...

Post your preference when you resolve your clog. Good luck!

--
fotografer

 

[Greg_E]

I'm going to throw another choice into your testing (I haven't used it yet). Someone on another forum suggested that this software can sometimes do things that the others can not, so you might want to look into it.

http://www.fixerlabs.com/New_Website/pages/sizefixerxl.htm

They also have a size limited version that is cheaper.

As far as the test goes, I would need to print each sample and then let my RIP upsample the original before I could play along. As one poster said, it's all (mostly) about the print. This water drop image would probably go very big without too much loss. I also ran acroos a technique where you split the image into the luminance and chrominance channels, then upsample the lum channel and somehow match the chroma back to this new size image. I'm not entirely sure how it works, but they claimed to be able to get very large percentages with this. It came from one of the digital camera back makers, and was to help you get your 6MP images up to a large print size. It all sounded plausible, but again, i did not try it.

--
http://www.dfaprinting.com

 

[The Davinator]

Here is a rough sample of what I was talking about. As some people have monitors at 72 & 96 dpi, this is just a guestimate.

I have reduced the samples you chose to approximately reflect what they would look like on a print vs screen. All three were reduced the same way. All three are untouched.....ie: no sharpening, etc. All three were reduced using bicubic only.

Now you'll note that the sample "F" that you chose being best appears soft. Compared to sample "0" (the original) and sample "A" (which you chose as worst). This shows how what appears on screen at 100% doesn't always show what is best on print. In this case, the sample you chose as worst (A), actually appears closer to the original.

This is why on screen, we have a lot of inexperienced people shouting "artifacts, halos", etc, etc, when in fact, on print, the same photo would look better. This is something gained from experience. Unfortunately, this often gets lost in the nonsense found in on line forums.







Best regards,

 

[fotografer]

Yes, Dave, I am not disagreeing with you. Thus, I qualify my choices by saying based on on-screen viewing (in their 'original' size, not the 'reduced' size you've shown here). Hence, I picked F, not A.

I have a pretty good idea how A will borne when printed... I have a sneaky feeling that when printed, my preference may well be the complete opposite!

--
fotografer

 

[Ron AKA]

In response to your post and Dave's further up, I appreciate that printing in itself, and printing at large sizes introduces even more variables into the mix. However, my purpose was to keep the complications to a minimum. So, I chose a simple image, with dimensions that fit into the default window of this forum so they can be seen basically the same by all.

So we could make large prints, but all in all I think this is a fair test with the starting point of 0.012 MP extrapolated up to 0.114 MP at 5.5" wide. So while these are not prints measured in feet, I think this is a significant stretch equaling about 10x in megapixels. Can you imagine trying to sell a 0.1 MP camera for 4x6 prints? That's what you are seeing here.

Ron

 

[Ron AKA]

Opps, I slipped a digit there. "Imagine trying to sell a 0.01 MP camera for 4x6 prints?"

 

[Ednaz]

Even qimage says that you should not even pay attention to what the screen image looks like. You only have to look at a screen soft-proof, or screen 100%, then print the same image and take a loupe to it, to understand that evaluating an image on screen is only useful if the image will only be used on screen.

There are several reasons, but the one that I think contributes the most to the differences is that on-screen images are luminous, where prints are reflective. Even color theory is different for light sources versus reflective surfaces; rgb versus cmyk. Then there's also pixels versus dots, dithering on paper versus on screen... the list is so long that I stopped paying attention after hearing the first two differences, and tried printing versus screen.

In terms of color and soft proofing for color fidelity, you can get awful close with a crt to paper. in terms of resolution, edge effects, and things like the effect of one paper surface versus another, there is no substitute for prints.

 

[Ron AKA]

I am sure that if I printed all these images with Qimage and my HP printer set at 1200 ppi input the results would be different again. However, what that is doing is introducing another variable into the evaluation of the interpolation calculation. The purpose of this exercise was to single out the interpolation method and compare them side by side without introducing printer variations. These images represent what the printer has to work with. I would suggest one would have to be going blind not to see at least "some differences". I guess the question is which one of these images would you send to your printer with the most chance of it replicating the original image?

I would suggest that the need to see a print is a red herring in this exercise.

Ron

 

[The Davinator]

I would suggest that the need to see a print is a red herring in
this exercise.

Ron

Considering that the whole purpose of interpolation is to output to a printer and not stare at a screen......I respectfully disagree with the above statement.

 

[Ron AKA]

So, Dave your choice is???

 

[The Davinator]

So, Dave your choice is???

A,B, & D are very similar for print output. I would choose A or B with this particular image.

 

[aruzinsky]

The
interpolation method used for both the downrez and the up rez were
the same.

That is unlikely because the terminology for reduction methods is ambiguous. For example, there are at least three ways to do "bicubic" reduction. The most popular method, and probably the best, uses a bicubic convolution kernel that is expanded in inverse proportion to scale. Another method will first use some low pass filter followed by an unexpanded bicubic kernel. Still another, the worst, would be to just use an unexpanded bicubic kernel. The point is that you can't depend on a name to unambiguously describe the size reduction method, particularly in cases such as QImage's pyramid which have no real reduction counterpart.

Also, even if the names were unambiguous, you do not want to see the effects from different reduction methods confuse the enlargement results. Thus, I would prefer to see just one reduction used as a test image for all enlargement methods.

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

 

[fotografer]

Ron,

I have done my own printing of these images. I saved them from your original posting, put them (all seven of them, including the original) side by side using Photoshop, at 600ppi (without uprezing or downsampling), so each 'splash' is tiny in size.

Then I use Qimage to print them all together at once at 600ppi using HP ps7150 at Best setting (HP native printer resolution is at 1200dpi, but at Best mode, only 600ppi is required without interpolation).

True to my (and Dave's) prediction, my preference of the printed samples is almost the complete reverse. Image F in this case looked the most 'soft', while image A looked the 'best' (sharpness, constrast-wise, closest to original). But I can still detect the usual 'artificial' accutance in image A (but not in image F, which is 'silky smooth', let down by the apparent lower 'contrast').

--
fotografer

 

[KCampbell]

The other three turn me off fairly equally because of artifacts.

Kevin

 

[Mike Chaney]

There are several reasons that you might choose one method over another in print versus screen. One major factor is final print sharpening. If you take an image that has been sharpened for the screen and dump that to the printer, it will not have the proper sharpness because sharpening at 72, 96, or 120 PPI for a screen will appear quite different when you print at 600 or 720 PPI. A smoother image with less artifacts can handle a lot more sharpening without falling apart, and the print process in general can often handle more sharpness than the screen without visible artifacts.

So a smooth interpolation algorithm coupled with a smart print sharpening routine produces the best results. This is why Qimage has smart sharpening for the final print that factors in things like how far you stretched the original (how much it had to be interpolated), the PPI being used by the print driver, and other factors so that the sharpness of the print is consistent and comparable to what you see on screen. The problem here, and the reason you might prefer F over A on screen but A over F in print, is because the samples in this thread have already been upsampled without factoring in the sharpening needed for the printer.

With this in mind, final print sharpening can be used to make F look better than A regardless of whether you are displaying on screen or printing. See the samples below. I took F and applied some more sharpening, and now, as you can see, F is just as sharp or sharper than A and will appear so on both screen and in print without the aliasing artifacts. So when looking at the samples here, you have to factor in both smoothness and sharpness as to which is better for what purpose. Also keep in mind that the dithering that is inherent in all inkjet printing often hides some artifacts like aliasing so that they are not as noticeable as they are on screen. For this reason, differences between interpolation methods can often be less noticeable in print than they are on screen. Of course, a lot depends on the printer, paper, and other factors as well.



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