David Pogue has faced his
fair share of myth-busting
duties, but his latest assignment sure didn't have to be done in order to convince us that his unanimous discovery was indeed correct all along. While theorists (and those adamant that bigger
always equals
better) can theorize forever on why
additional megapixels should yield clearer, more accurate prints, the reality of the matter is that
extra megapixels are typically not much more than a
marketing ploy to lure consumers into making an additional purchase. In his latest test, he took identical photos with anonymous 5-, 8-, and 13-megapixel shooters, and then printed them out on 16- x 24-inch poster paper at a professional photography lab. Then, he surveyed the general public in Times Square to see if the naked eye could actually discern between the varying sensors and the level of clarity -- to no surprise (and to theorists' chagrin), only a single person (a photography professor, mind you) correctly sorted the prints in order of megapixels, and Pogue even asserted that the lucky winner was probably guessing anyway. So, if you still don't believe us, be sure to hit the read link for the full skinny, and save yourself a few bills the next time you're camera shopping by not making megapixels your ultimate priority, capishe?
[Via
TheRawFeed]
Reader Comments (Page 1 of 1)
Michiel @ Nov 23rd 2006 4:54AM
Are there still people who believe the megapixel hype, right?
When shopping for a camera, first consider the way you intent to use it.
Point&Click? Think about shutterlag, reach of the flash, ease of use, F5.6 lens means that you will have to start flashing even before dusk. Do not buy a Fixed lens auto-zoom monster, or SLR. You're throwing away your money.
Enthusiast? What kind of shooting are you going to do? Know that the lens you got with your SLR is shit. Better to buy only the body and spend a lot of time reading about lenses. Realize that flash will ruin your photo. Realize that because of this you will need at least a F2.8 lens. Consider fixed-focus lenses, they are faster.
Pro? No need to explain, you already know all this :)
steve miller @ Nov 23rd 2006 9:13AM
I don't think Pogue understands that most people fail to fully use the frame and wind up needing to crop their photos a lot. If you crop out half of a 5 meg frame and then try to blow it up to 11 x 14 your left with a poorly detailed photograph. As a professional I sometimes need to pull a single person out from a two person side by side photograph which winds up cutting my 10 or 12 meg cameras resolution in half and the images still hold up well. If I had to do that with a 5 or 6 meg camera the images would not look so good. Then take a situation where a person is in a group of people and it is the only frame you have of a person who is suddenly a headline for a story. Now I might have to use only a small portion of the image. This has happened many times. Pixels do count. If resolution didn't matter then Ansel Adams wouldn't have bothered using an 8x10 view camera or have written many times about how important a quality negative was to making a great photograph. The pixel count within a raw file is a vital link to a great photograph. If all you want is to make 4x6 snaps then any 3 meg will do. But if you want to make good photographs it helps to start with a well detailed file.
cjrenaud @ Nov 23rd 2006 10:18AM
You obviously haven't used the kit lens with the Nikon D80, then. Find a better lens out there for under $700. I'll wait here with your plate of crow.
Gil @ Nov 23rd 2006 4:56AM
Who gets their tech dose from the NY Times?
Michiel @ Nov 23rd 2006 4:58AM
Forgot to mention my point:
All cameras have sufficient megapixels for their use by now. Don't even look at it.
Megapixels are like processor megahertz. No, because we now all understand that memory, your videocard, soundcard, broadband bandwith, etc. are all much more important (now that processor megahertz have reached the "sufficient" level).
Adam @ Nov 23rd 2006 4:58AM
Uh, I've always noticed a marked increase in quality when comparing a lower resolution to a higher resolution picture. This obviously has more to do with the quality of your printer than the quality of your camera. Seems like a flawed experiment to me.
xVariable @ Nov 23rd 2006 4:59AM
If you want to do large image sizes, you need big megapixels. Of course noone's going to notice a difference at 8x10 just like noone's going to notice 1080p at 32".
suntiger @ Nov 23rd 2006 12:09PM
I would.
Nogami @ Nov 23rd 2006 5:08AM
"mexapixels don't matter" is incorrect. What he should be saying is that it's a "diminishing return" for consumers. After a certain point, the increased resolution won't improve the picture in a direct correlation with the increase in the cost of the camera.
For professional cameras, there are also other considerations - a 6Mp professional DSLR is going to blow away a 6Mp consumer point-n-shoot every time, but that includes better optics, better processing and capture code, etc.
It's also worth noting that his sample pictures are all baby photos against a neutral background. Things would likely look a lot different if the photos were composed of objects and people at widely varied distances from the lens, under a variety of indoor and outdoor lighting conditions.
Bobby Johnson @ Nov 23rd 2006 5:12AM
hmmm? not wanting to pick up the glove for the (evil) corporations it should be noted that the count of pixels maybe doesn't have anything to do with the quality per say but a higher pixelcount does give the possibility to make larger prints. if all three photos were printed out in the same resolution, they will of course all look (pretty much) the same printed in the same size. there's no detailed info on how the photos were processed (he probably had to blow up the 5MP and maybe the 8MP pics using imaging or printer software) but at a first glance it's not really a fair test. shoot me if i'm wrong.
Jeff @ Nov 23rd 2006 5:27AM
Re: whoever was talking about SLRs. Kit lenses aren't that bad. It's the photographer, not the equipment. If the manual settings are there I can take perfectly good pictures with a cheapo kit lens. Would I rather have my expensive lenses? Sure. But they're not completely necessary unless you're a talentless gear whore.
Ansel Adams @ Nov 23rd 2006 2:04PM
Oversimplification.
Most kit lenses are great if you are not very demanding, not 'talentless'. I suppose if you don't like accurate colour reproduction, you're a fan of autofocus hunting, not being able to shoot in low light turns you on, and a crisp image seems unnecessary, a kit lens may be for you. For everyone else, drop it and pick up a real lens. Anyone who tells you that you won't notice a difference is blind. Try a nice 50mm f/1.8 fixed lens for about $100 and tell me I'm wrong.
As stated in the article, megapixels are important, but not when limited by optical resolution, which is the case with most digital compacts. Believe me, there is a reason why a 22MP digital back costs tens of thousands of dollars. Look at a billboard along the highway on your drive to work.
alexq @ Nov 23rd 2006 5:28AM
Just playing devils advocate here folks ;=)
The test is flawed because it was not double blinded. Shrugging off the one person who got it right as probably down to luck is just the testers opinion and not based on any evidence.
Sure, there may be little difference at the end of the day in visual quality, but it doesn't give much credibility to the result if the tester is biased and shows this in almost discrediting the one person (a "photography professor") who got it right!
What's needed is a proper test conducted independently without influence from those who have an agenda either way.
3rdsun @ Nov 23rd 2006 5:39AM
How many magapixels is that picture taken with. Seems like 0.1 pixels
Carniphage @ Nov 23rd 2006 5:50AM
One of the seven deadly geek sins is .... Sin Number 2 - "Comparison performed via mathematics".
It works like this - In order to compare two things, look at the stats. Megabytes, megahurtz, millimetres, microseconds. etc. See which has the better numbers... and bingo. We have a winner!!!
Thing is, that ignores all the really important stuff, like experience, aesthetics and perhaps the most important of all: outcome.
I could cite examples. But I feel a billion flames would burst into existence.
Vince @ Nov 23rd 2006 6:00AM
ahh. interesting actually. They "down-rezzed" the photo from 13 to 8 then 5MP, and printed it at a pro photo shop. So, if 95% of the people taking the test gave up and couldn't tell the difference, it kind of makes you wonder if that pro gear doesnt do some kind of up-scaling.
kadajawi @ Nov 23rd 2006 6:01AM
Well, if among a hundred persons (they should have said how many were tested) only one got it right it's rather a proof that the higher MP can mean worse quality. But yes, the test is a bit flawed.
@Nogami: Why should it? You would obviously notice a difference if the sensor has a different size, the photo was shot with another f-stop, etc. But that has nothing to do with resolution.
paul @ Nov 23rd 2006 6:15AM
All this shows is that the "professional" photo-lab where he got his prints was crap. Do it yourself... Take a photo with ANY digital camera, print it, downscale it and print it again at the same dpi - the results are glaringly obvious. Even better someone do it and scan the prints at the same res so we can all do this "test".
Vanillacide @ Nov 23rd 2006 6:18AM
This test is technically flawed in two ways:
1. downscaling is not appropriate. A 5MP image downscaled from 8MP will look better than an image shot at 5MP; multi-tap interpolation will produce a smoother image, for example no banding.
2. limited by quality of print, not enough resolution to show differences.
In practical terms though, he could be right. For most consumers megapixels do not matter that much.
However, if you want to zoom and crop then you need the additional pixel resolution to get the same quality results -- unless you have the magic pixel machines that CSI use on television to solve cases.
Whynot @ Nov 23rd 2006 6:20AM
Although I do agree that for the average consumer, 5Mpixels is enough to print in 8x10", his test is flawed. He downsized the pictures, therefore cancelling the difference between the cameras, which makes the test useless.
"Let's make a test with a 5, 8 and 13Mpix cameras, but we'll downsize all 3 pictures to 1024x768 and try to see a difference" That's stupid! If you downsize the picture, you lose the only thing that truly made the difference.
The point in having a 13Mpix camera is that you can have a 4368x2912 sized image. Of course it's not usefull if you print on 8x10", but for professional use it does make a difference.
Pointing out that 5Mpix is enough for most people is fine. Tricking people into thinking there is absolutely no difference between 5 and 13Mpix is not.
Joe @ Nov 23rd 2006 6:27AM
Megapixels do matter, but not when they are jammed onto a tiny sensor as in most every point-and-shoot digicam. Laws of physics get in the way and the inevitable noise and lack of sensitivity limit the advantage of adding more and more pixels in the same space. In other words, there's a difference between the quality and the quantity of the pixels.
In DSLRS, the pixel pitch is far greater and there's a noticeable difference with more resolution, but not near as much as people might expect with each small jump (6 to 8, 8 to 10). In reality, you need four times the megapixels to effectively double the resolution.
And, like someone mentioned above, a portrait is not really the best way to compare resolution. Take a good landscape with lots of small detail and the differences become more obvious, especially between most small sensor cameras and a dslr.
It's just sloppy journalism.
Kevin @ Nov 23rd 2006 6:30AM
The test is flawed because the photos were not taken at the two lower resolutions. They were taken at the highest resolution (with the implicit admission that this would give the best image), and then scaled down to the other two. The write-up above falls into the trap of assuming that different sensors were used, when in fact the (theoretically) highest quality was the only one used.
I'm presuming that they used photoshop or something like that which would use some good scaling algorithms, and its 5MP version of a good 13MP image is likely to be better looking (NB, I'm not saying it's better quality, just better looking) than a native 5MP version of the same scene. Plus, the resampled images would also be likely to scale up a bit better to the print size they used, as the edges would have been anti-aliased (a proper 5MP image would usually have harder edges which would be more noticeable in large print sizes). Also, given the subject matter of the image, people would be more likely to favour a softer tone - The whole thing is flawed in so many respects that it's hard to look at it as anything other than an exercise in pandering to a certain demographic (e.g. see some of the comments on the page linked-to). I don't know who this David Pogue is, and stuff like this certainly doesn't make me want to find out.
Personally, I believe that when you get up to 5MP, the optics start to become the most important component, but 'experiments' like this do nothing to bolster that opinion when the details reveal their lack of credibility.
Jeff Lewis @ Nov 23rd 2006 6:33AM
Uhm... to all the people arguing about the *camera's* resolution - let me point out the obvious - it's the *printer's* resolution that determines the maximum quality.
Everyone in the printing industry knows this - heck, it's hard coded into Photoshop's resize image panel.
Nyquist's Law says the maximum sampling rate to acheive a perfect reproduction of a repeating wavefunction is 2x the highest frequency. In the case of printers, the max frequency is the dots per inch. So the maximum useful resolution for the camera is twice this.
But wait - if it's a CMYK printer (as most are) then you can't create just any colour - you're limited to essentially 10 colours. To make up all the rest of the colours, you group the page into cells and use all the dots within the cell to represent the colour. The standard cell is 6x6 (although different patterns and cell sizes exist).
That means the effective resolution of the printer is 1/6th its real resolution. Double that (see above about Nyquist Law) and you get a max required resolution of 1/3 the printer's real resolution.
So, if you have, say, a 600dpi printer - the effective resolution is 200dpi across say 8x10 gives you 1600 x 2000 max resolution.
In other words, anything larger than 1600x2000 will give you NO additional quality.
The only exception is dyesub printers like the Selphy - where you need 1:1 (it doesn't dither) - but you'll also find the printer's resolution is surprisingly low - 215dpi is typical - because there's no need to blur the dots - your eyes do that for you. So the max resolution still ends up about the same.
Bigger is not always better.
Peter @ Nov 23rd 2006 9:25AM
You are mistaken about modern printer technology. It is no way a limitation.
Typical Specs:
Advanced Micro Piezo 8-color pigment ink jet technology, optimized for photo printing
Variable droplet size. Minimum 1.5-picoliter
5760 x 1440 optimized dpi
Easily capable of bettering the limits of vision (~300dpi). Vision will give way before any decent photoprinter. At 300 dpi with 16"x24" paper the print would only start to be the limiting factor around:
16*300 x 24* 300 = 4800 * 7200 = 34 MP.
The print is not the problem, the chosen source and the downsize are.
Ed @ Nov 25th 2006 10:47AM
Jeff, what determines the quality is the actual size of photosites too, you could have the best printer around but if your camera's sensor has smaller photosites it will loose in aquiring more details, dynamic range and it will be prone to noise...
The combination of a great printer that you describe and a camera with a decent sized sensor is what will fare best for a good photo (in technical image quality side, not photographic skill of course)
TorontoGuy @ Nov 23rd 2006 9:04AM
Any of these "and nobody could tell the difference" tests are irrelevant and basically dishonest. These 'tests' are like optical illusions...they only fool you if you don't know the trick and, after that, you can see it everytime. In this case, the 'trick' is the old resolution 'shell game'. With a confusing mix of upsizing and downsizing...resampling and resizing...there is nothing that can be concluded except that the tester is pulling a fast one.
Peter G @ Nov 23rd 2006 9:09AM
All this shows is how little typical mainstream reporters understand technology. Even the "technology" reporters.
The first mistake is in using a downsized image. Digital cameras you the Bayer CFA to produce color, which they have a sparse amount of sensor info. They use an algorithm to determine the missing colors. The net effect is that you can downsize a typical bayer image by half an essentially lose no details, because the size of the output is something of a misrepresentation. So that 13MP bayer source image is in reality only a container for about 6.5 MP of information. Even at his smallest downsize of 5MP there is not a significant difference. A real 5MP digital camera would only contain about 2.5mp worth of information. There would be much more drastic difference using a 5MP camera, than a 5MP downsize which is really equivalent to 10MP digital camera output.
Second major flaw was choosing a largely low detail subject. Not a subject where people expect/want high detail.
Repeat the test with a detailed landscape shot and using actual cameras instead of a downsize and then he might have something to make a case.
As shown, this is flawed evidence either due to shallow understanding of the issues, or purposefull misrepresentation.
I may agree that megapixel madness has gotten out of hand, but this is a completely faulty test, that proves nothing except the limited knowledge of the tester.
Rick Lyon @ Nov 23rd 2006 9:30AM
THe majority of comments here seem the neglect the target and reason for the test. Does my mom need to spend the extra $150 to get the 10megapixel camera or will the 4 mega pixel camera be good enough? That's the whole point here. BEcause after all, MOST of those who buy one of these cameras IS NOT A PRO and will AT MOST be printing out 8x10 photos! So, anything over 4 meg pixels is a waste of money FOR CONSUMERS.
I've been ranting for 2 years now, instead of mega pixels they need to improve zooms. 3x zoom is still the sad standard.
cjrenaud @ Nov 23rd 2006 10:29AM
Rick, you missed the point. He didn't take a picture with a 5MP camera, then take another picture with a more expensive 8MP camera, then another with an even more expensive 13MP camera.
He took ONE picture with the 13MP camera and downsized it to 5MP- and 8MP-resolutions. That will always look better due to interpolation.
And, as has been pointed out before, he should have used a subject that'd be more representative of how people use their cameras day-to-day. A picture with the family in front of the Christmas tree opening up toys (where you'd have a lot of detail at different depths and a variety of colours). Or, even better, the aforementioned landscape shot.
shmengie @ Nov 23rd 2006 9:41AM
yeah, but where's the sweet spot? does that mean that my 1.3mp camera is just fine? or do i need 3mp? 5mp? 6mp? when is it too much?
Rasmus @ Nov 23rd 2006 10:33AM
Well as an amateur, you will probably show your pictures to family and friends on a huge TV or on a projector. That is probably the worst case scenario, so to speak. The latest standard for TVs and projectors is 1920x1080, which is 2.1 M pixels...
mikemobility @ Nov 23rd 2006 11:28AM
I've only briefly skimmed the comments above, but I didn't see anyone mentioning the other disadvantage of the megapixel race: CCD/CMOS real estate and pixel size.
I believe most, if not all, point&shoot digitals are still using 1/1.8" or 1/2.5" sensors (and most of them made by Sony too). To the best of my knowledge, as they've gone from 3 to 6 to 10 megapixel compacts, they haven't done much with the sensor size either, so what you get is more photosites in the same space, which means smaller photosites which affects a couple things:
1. less light per photo site, affecting tone, sharpness, etc.
2. higher temperatures on the CCD surface = more noise.
While a previous poster is correct that a 6MP DSLR will blow awway a 6MP P&S because of the optics, it's also because of the sensor size. 6MP spread out over the APS-C-sized sensor vs. the 1/1.8" sensor makes a big difference.
While not scientific or conclusive, it's illustrative just compare even a Canon 20D/30D's 8MP images @ 400ISO/800ISO to the equivalent Canon SDXXX @ same res, same ISO. Same argument goes for making the jump from APS-C sensors to Full-Frame (35mm) sensors.
greg3d @ Nov 23rd 2006 12:41PM
Were the images printed at 150dpi? 300dpi? 600dpi?
It makes a big difference on how much detail you get in the image when printing a 5 vs 13 megapixel image at 600dpi.
Also, downrezing a 13mp image to print in the same format as a 5mp image would naturaly produce similar results. A tech writer should know better.
kadajawi @ Nov 23rd 2006 12:49PM
Ok, now I've read the actual blog item, that was just stupid to downrez the same picture. Of course that will yield a different result than using 3 different cameras.
Yes, if you crop alot, if you want to print posters, then you're going to need a lot of resolution. But most don't, I think 5 MP should be more than enough. And more is not always better, since usually the more MP you have at the same sensor size, the more noise you will get. Which is very bad. Low light performance is affected too. So I'd say 5 MP for a point and shoot camera, if it's a DSLR 6 and more makes sense.
Taking a photo with subjects at different distances may reveal the speed of the lens (or better how the photo was taken) and the sensor size, but otherwise I don't think it makes alot of sense. The bigger the sensor and the faster the lens, the less depth of field you will get.
v_dogg @ Nov 23rd 2006 12:51PM
lol i like how the picture is super low res.
helio9000 @ Nov 23rd 2006 1:20PM
Boy Pogue is just terrible. As usual it hard to know where to begin in picking out the flaws in his logic. I haven't ever learned anything new reading his stuff (unless it later turned out to be wrong). In a typical column he is either stating the obvious or spreading simple-minded misinformation. A dual action threat.
The problem is that he has a big platform and the only people who turn to him seriously for tech news are those least able to make sense of his ramblings.
StickMaker @ Nov 23rd 2006 1:35PM
Michiel was right-on when he said: "When shopping for a camera, first consider the way you intent to use it."
I bought a very expensive SLR Canon to photograph wildlife for photos on my website at http://www.sticksite.com/ FIRST picture proved I'd goofed; the shutter noise scared off the deer. So, I got the Panasonic DMC-FZ20; no more noise; no more problem.
The books don't tell you that.
Jacob @ Nov 23rd 2006 2:51PM
"Megapixels don't matter" is complete and total nonsese... unless all you have in mind is "point, shoot, and print".
Me? The more pixels I get, the better. The reason being that I like to be able to go into my photos and crop-out little bits and pieces. The more visual data I have to work with, the better quality result I get from cropping-out small bits.
Rod @ Nov 23rd 2006 3:08PM
Before you comment on this topic, if you don't even own a DSLR, please don't.
max andrews @ Nov 23rd 2006 3:09PM
Yeah this test is BS. I work in a large format printing facility and pixels matter very much, especially at big sizes. The fact that he down-rezzed each photo to the same number of pixels is retarted. Of course you won't be able to tell. In fact, ALL of the prints definitely looked like crap. Given that if you are a foot away from a print you can easily see the individual pixels in even a 150 ppi image (he also confused dpi and ppi, which certifies he does not know how to do this test properly), 16x24 at 150ppi equals 8.6 million pixels. This is about the threshold for a decent print at this size, and as such you NEED at least this many pixels. If you down-rez a 13mp image to 5mp and print it's going to look just as crappy as the other native 5mp image at 16x24. What you need to do is print a 16x24 image at 13mp and the other at 5mp with no interpolation (upscaling). If you can't tell the difference between those two you need to be maimed.
THis article is trying toconvince you that you can't tell the difference between standard definition and highdef. And that's simply not true.
kev @ Nov 24th 2006 1:03AM
16-by-24 is not big, but it is bigger than what you would normally print out. Nevertheless, can he just shut up?
kev @ Nov 24th 2006 1:04AM
By the way, I bet this guy doesn't understand the difference between digital and analog without looking up their definition in the dictionary. It's always good to have more megapixels. Hell, it is said that the human eye is capable of hundreds of megapixels.
Ed @ Nov 25th 2006 10:47AM
What people should learn is that the size of the sensor is what matters not the megapixel count...
An APS-C 6 megapixel sensor will be leagues better than a 10 megapixel little finger nail size sensor... and an APS-H or full frame sensor would beat the crap out of the first two mentioned!, photosites size will determine the final quality of the final photo! also diffraction will damage you more with smaller photosites!!
In this XMAS don't fall in the trap of more megapixel compact cameras... if a compact camera has more megapixels it only means more noise, less details and problems with diffraction (when the aperture is stopped down)!!
Hugh Fulljames @ Nov 25th 2006 12:25PM
I think resolution does matter - the reason - picture cropping.
If you are just going to take an image and print it out then fine, I agree 5 megapixels will is probably enough for most output requirements.
However, if you decide not to use the entire picture but to crop in on, say, a face then the extra resolution is going to be useful.
Seth @ Nov 27th 2006 9:32AM
The other issue is megapixels vs usage.
As digital photography becomes more popular, printing is not as much of an issue, as 99.9% of photos will never be printed but instead viewed on a computer and emailed or flickrd.
In this light the extra megapixel storage is largely wasted, and require the images to be shrunk before sending or posting.
naphra @ Dec 14th 2006 9:57PM
I always thought that the theorists always knew this, and that it's the consumers who thought that megapixels matter.