Welcome to Photo.net: A Community of Photographers

Other items:

View Photo.net Holiday Gift Guide »

Pixel Peeping Part I

By Bob Atkins

Since digital cameras (and especially DSLRs) first became popular there has been an epidemic of what might be termed "Pixel Peeping". Pixel Peeping is a secretive activity practiced by some photographers in their basement and involves repeated photographing of resolution charts and grey cards, followed by manipulation far beyond the bounds of normal in an attempt to find defects in the image.  It's sort of like giving a hypochondriac their own magnetic resonance imaging machine and finding them spending all their time laying in it trying to prove that there's something wrong with them!

Now of course equipment testing is an old and well established tradition, and indeed it's something we all do at times. If some thing is wrong, it's good to know about it. The problem is that if you look hard enough and you don't really know what you're looking for, you'll probably find something.

So here's a Pixel Peeping Problem. Take a look at these two images. I've been testing quite a few cameras and lenses recently, both for photo.net, my own website and my own information. These are crops from a couple of images I shot of the same resolution test chart during the course of these tests. I'm not going to tell you anything else. I'm not going to tell you which cameras, what type of cameras, what lenses or anything else. All I'll say is that the images are unmanipulated and that they were shot and cropped to have exactly the same image scale so as to appear as identical as possible. I will tell you that I'm not "cheating" and showing the same image with different post exposure processing. They are crops from two different "out of the camera" images.

peep_a_b.jpg (30300 bytes)

What would you conclude from peeping at these pixels? What would you say about the cameras and/or the lenses and/or the camera firmware?

In "Pixel Peeping Part II" I'll give you the full technical details on these images so we can see how accurate your peeping was!

 

© Copyright 2004 Bob Atkins (www.bobatkins.com)

Readers' Comments


Add a comment



Bob Atkins , December 05, 2004; 04:40 P.M.

Please excuse the "Best Lens for a Leica" title you may see when adding a comment. It's a system glitch (my fault) which I'm trying to correct. Your comments will appear on this page though.

I'll just add that these images were not shot with the same camera using different "in camera" settings either!

Struan Gray , December 05, 2004; 04:57 P.M.

The first is from the new Sinar 22 MP back used with a 100 mm Schneider APO-digitar HM lens. It shows the finest set of bars on the chart and indicates a system resolution of 243.2 lpmm. The second is a Gameboy camera image of the coarsest set of bars and indicates a system resolution of 0.23 lpmm.

I should of been a lawyer.

Ilkka Nissila , December 05, 2004; 04:58 P.M.

It looks like the top image was taken with a camera without an antialias filter, while the bottom pic was taken with an antialias filter in place. How many points do I get? ;-)

Guy Hammond , December 05, 2004; 05:51 P.M.

Secretive? On most forums (see dpreview.com for an example) pixel peeping is practically all they do...

Roger Krueger , December 05, 2004; 06:00 P.M.

Image A is from a high-megapixel but cheap P&S that has no antialiasing and gonzo sharpening and contrast boosts.

Image B is from a high-end DSLR with good anti-aliasing, and with in-camera sharpening and contrast boost turned off.

Jon Austin , December 05, 2004; 06:12 P.M.

I'd say, based on the images shown, I'd take camera B over camera A any day.

Carl Smith , December 05, 2004; 06:36 P.M.

A is a Kodak 14.5 mp camera, B is a Canon 1Ds.

There's my guess. :)

Charlie Vigue , December 05, 2004; 07:10 P.M.

Yep, top (A) looks like it's showing aliasing artifacts, false detail, whatever you want to call it. Sample 'B' looks softer but that can be well corrected in post processing.

I'd guess sample 'B' is pretty close to an optimal AA filter due to the beginnings of what look to be 'false color' in the lines immediately higher than the spatial resolution limit. Or I'm seeing things. :-)

Niels de Boissezon , December 05, 2004; 07:15 P.M.

Exhibit A is pretty much impossible to process any further as it obviously has been USM'd a lot already. Exhibit B has very little USM to it and can most likely be massaged to look better than A.

My guess would be a 7-8 MP digicam (A). And a 8 MP DSLR (B).

Kim Olsson , December 05, 2004; 07:32 P.M.

Pixel Peek A looks like it comes from a D70 with its weak anti-alias filter. I've seen it once in over 2,500 pictures taken with it.

I guess that Pixel Peek B is taken with a Canon D'ish camera.

Lens? I have not the foggiest...

Elliot N , December 05, 2004; 07:44 P.M.

A, Nikon. B, Canon.

Dan Chabanov , December 05, 2004; 08:39 P.M.

My guess:

...

This is completely and utterly stupid!!

...

This article contributes zilch to anyone who is interested in taking photos of anything other then a resolution chart.

Alistair Windsor , December 05, 2004; 09:10 P.M.

There are so many variables to consider. Many of us using the new Canon software have noticed that the ame RAW shot looks different when processed by different converters. Canon's new DPP software for example exhibits significantly more moiré than Canon's EVU software.

John Schroeder , December 05, 2004; 09:36 P.M.

My guess is; Image A was taken with a film camera and then scanned in at a resolution to match, as closely as possible, Image B which was taken with a digital camera. The same lens, apeture, and iso were used on both cameras.

Beau Hooker , December 05, 2004; 09:53 P.M.

Without meaning any disrespect to Bob, but - what is the point in guessing? Keeping the peepers busy, I guess. This reminds me... Good old vinyl, LP records back in the '70s many times had tiny messages scrawled in the vinyl itself, concentrically around the center of the album. I recall one that said "How do you keep a Black Sabbath fan busy? Over . . ." When you turned the album over, the same message was on the other side. ;-)

Bob Atkins , December 05, 2004; 10:01 P.M.

The point in guessing is to show just how many different interpretations you can make - most of them reasonably consistant with observations - given the same data.

It's clear that even when people do the same tests, they can draw different conclusions from them. Person #1 says a lens is great. Person #2 says it's not. Who is right and why do their opinions differ? People see all sorts of things, even when they're not there.

Shown these two images, equally skilled and knowledgable people will come to different conclusions.

Is anyone right yet? I'm not saying...

Part II of this article might make things (and the point of all this) somewhat clearer. Then again, it might not!

Ryan Joseph , December 05, 2004; 10:14 P.M.

Hmmmm.... option A is probably from a consumer camera of some kind or the Kodak DC 14/n or whatever its called. Option B is a relatively unproccessed image from a DSLR with a CMOS sensor I'd guess. Or Bob could be pulling our legs and these could be from the same camera and just processed diffrently....

Bradley Phillip , December 05, 2004; 10:24 P.M.

Same file, different raw converter?

If so, there are lots of implications that don't even have to do with the camera or lens hardware.

Yes, testing equipment is complex. But some of us care and some of us don't.

To each his own, I say...

Jamie W , December 06, 2004; 02:06 A.M.

I say... who cares

No offence Bob :)

David Vatovec , December 06, 2004; 02:35 A.M.

Never shoot resolution charts, so i could not guess,...

The B sample seems a Canon DSLR to me.

James Davison , December 06, 2004; 03:12 A.M.

Pot...kettle...black.

Yoni Shveka , December 06, 2004; 07:57 A.M.

I would guess A is from Digital Rebel and B from Black Bodied Digital Rebel (why else would they charge 200$ extra for it?)...

John Crowe , December 06, 2004; 10:14 A.M.

Although the top image is showing some horrible "artifacts", if I may use that word, (I am not quite up on all this digital jargon), I believe in real world picture taking this would produce sharper and more contrasty images.

On the other hand if the bottom photo was taken with a good camera and a $100 zoom lens then a good lens could go along way to improve the results.

As someone who is trying to learn as much as possible before going digital in the next 3-5 years, I am looking forward to the results of your little puzzle.

I don't mind this, "blind taste test". I recently took my lens tests from about 20 different lens/camera combos, with labels on the front of the slides, through them upside down on the light table and put them in order of sharpness as best as possible (about 80 slides). Then I flipped them over to record the unbiased results - it made quite a difference - it forced me to "face the brutal facts". (I put that in quotes since it comes from the book, "Good to Great" by Jim Collins).

csab' józsa , December 06, 2004; 10:17 A.M.

I don't know what cameras were used but these must be the two most boring images on photo.net, including my own port-foolio ;)

Mark Smith , December 06, 2004; 10:21 A.M.

What a waist of time, money and energy!!!

This is photography, not physics. Where is all this BS going anyway? Is it going to help you in your photography.

I would start with the grey pixel between your ears first!!!

M

Mark Smith , December 06, 2004; 10:21 A.M.

What a waste of time, money and energy!!!

This is photography, not physics. Where is all this BS going anyway? Is it going to help you in your photography.

I would start with the grey pixel between your ears first!!!

M

Ryan Joseph , December 06, 2004; 11:19 A.M.

In Bob's defense I think he is making a VERY valid point by posting this article. Bob is proving that there are NO absolutes in lens testing, and honestly I have trouble listening to people who decry lens comparisions as a waste of time, or who say the lens does not matter. Come on guys, at SOME POINT understanding these things is important. I doubt ANYBODY who has a choice between two 300 dollar lenses is going to choose the inferior one of the two!

Bob Atkins , December 06, 2004; 12:04 P.M.

Thanks for your well thought out comments Mark, though posting them just once would have been sufficient. Always good to know my efforts are appreciated. For the sake of your health I suggest you avoid reading Part II of this article, since it appears that it's really going to annoy you a lot.

Well, I have to go now. So many different resolution charts to photograph, so little time to do it...

Maury Cohen , December 06, 2004; 01:14 P.M.

I'll bet they're the same camera.

Scott Stadler , December 06, 2004; 01:23 P.M.

I going to say that the top one is the Canon 20D, and the bottom one is the 10D and he is illustrating that more megapixels isn't necessarily better, but just contributes to more noise.

Bob Atkins , December 06, 2004; 02:20 P.M.

The second shot would look exactly like the first. The differences are not due to any extrinsic factors like magnetic or electrical fields. I'm sneaky, but not THAT sneaky.

Maybe I should start selling framed prints of resolution test charts. Image "A" has some nice color effects. With a bit of cropping it might pass for art in some circles.

I'm definately sensing a degree of bigotry and prejudice against Pixel Peepers. This is wrong. I think we should at least try to treat them like normal people...

Jean-Baptiste Queru , December 06, 2004; 02:53 P.M.

Well, with little doubt the frequency response of "A" is higher than that of "B", throughout the frequency range (higher contrast on the left, response with bayer de-matrixing artifacts around Nyquist, Moire at higher frequencies. I'll claim that as a fact. Beyond that I'm having a hard time finding facts (especially as I'm away from Photoshop). Framing, exposure and color balance look very similar.

If we assume that the equipment is different (e.g. that the difference isn't purely the technique), I'd tentatively conclude that "A" may be more appropriate for most general photography, but that "B" might be a preferred solution for fashion and/or architecture where the strong response at and above the Nyquist frequency could be annoying against patterns in textiles or on buildings.

Now, knowing that Bob is a smart and sneaky person, I'd imagine that there's a trap somewhere. I wouldn't be surprised if Bob wanted us to try to guess which equipment was used for those two pictures, like some have done (guessing Kodak vs. Canon or Nikon vs. Canon). My guess is that the equipment is exactly the same, and that the difference between "A" and "B" is a difference in technique that causes "B" to be less sharp. I wouldn't be surprised if the difference was in focusing, or in using mirror lock-up (with a camera in portrait mode so that mirror vibrations affect the horizontal frequency response), or turning image-stabilization on and off. That's a wild guess, based not as much on fact as on a guess that Bob created on purpose an unexpected result.

[disclaimer: this comment was edited for clarity, and therefore isn't located in the thread in the same place as it was originally]

Bob Bell , December 06, 2004; 03:18 P.M.

Since another group has emerged out of photography, can we have claymation death matches between Pixel Peepers and Chimpers?

Marijn Bezuijen , December 06, 2004; 05:17 P.M.

Let me first stress that the conclusion I've arrived at after staring at these images for about a minute is an educated guess at best.

Unlike the guesses made by some other people, I do not believe these images come from cameras made by two different manufacturers. The color cast and overall image properties are identical. Both exhibit the same kind of color aliasing (a greenish area followed by a reddish area to the right of it). Also, the sharpening artefacts are almost identical, although slightly more apparent in the top image. This leads me to think that these images were taking using exactly the same camera (and at the same settings, as Bob already indicated).

So what would explain the differences? The images presented to the imaging devices in the camera must have been different in this case. More specifically, to take image 'A' a higher resolution image was presented to the imaging device than to take image 'B'. This higher resolution caused the highly visible aliasing.

One explanation for this would be the camera was equipped with an anti-alias filter for image 'A', but not for image 'B'. However, I've used a camera with a removable anti-alias filter and the differences were much larger.

So what I'd say is that the higher resolution of image 'A' was caused by differences in the lens. Either image 'A' was taken using a sharper lens, or the same lens was used at more optimal settings (medium aperture for 'A' versus large/small aperture for 'B', for example). I don't think there's a way to figure out the difference between those two scenarios.

Bob also specifically mentioned that the difference could be in the lens a few times in the original posting. I would be very pleased if it were in fact the case that image 'A' was taken using a better lens and otherwise identical parameters. This would not only demonstrate how more expensive (and theoretically better) equipment is not always better in practice, it would also prove how subjective and misleading even tests using resolution charts can be.

Jean-Baptiste Queru , December 06, 2004; 07:49 P.M.

Reading Marjin's comment, and staying along the lines of my thought that this might be the exact same equipment, the difference could also be in the aperture setting. A fast lens used wide open, or any lens stopped down too far (e.g. f/1.4 or f/40) would likely not be as sharp as e.g. the same lens at f/8, and the difference may be big enough to cause those two pictures to be created.

Oh, stupid extreme case, one shot was hand-held and the other one on a tripod (all els being equal).

Ken Dunn , December 06, 2004; 08:59 P.M.

Same camera and lens, just different processing, IMO, just to prove a point

Bob Atkins , December 06, 2004; 11:08 P.M.

OK. Part II has now been posted!

Ron Chappel , December 09, 2004; 03:56 A.M.

"....or in using mirror lock-up (with a camera in portrait mode so that mirror vibrations affect the horizontal frequency response)"

Good call that one!
Also some interesting theories posted by others.I'll go checkout part II now-i hope the answer is interesting after all :)

Steve Kuan , December 09, 2004; 07:26 P.M.

Exhibit A is a digital photograph of something.

Exhibit B is also a digital photograph of something.

Therefore A and B may or may not come from the same thing that has been transformed into the digital binary domain.

;)

Landrum Kelly , December 12, 2004; 10:27 A.M.

"This is photography, not physics." Well, photography is about optics, and optics is a branch of physics. Forget that at your peril. Of course, if you don't want to understand what is happening, just set the camera on "P" and snap away. You might not even notice the difference.


Add a comment



Notify me of comments



Photography