The New York Times' David Pogue has written the story that I've wanted to write for about three years now but never got to (mostly because consumer tech isn't my primary beat). Pogue has very articulately pointed out that there's a point of diminishing returns when it comes to the number of megapixels digital camera vendors are trying to sell you. The manufacturers in this space are crawling all over each other to you offer you more megapixels for less money. Chances are though, like the car that does 0-60 in 4 seconds (something you'll never really need in reality), having something in the neighborhood of 5-7 megapixels will suit everyone but the pros and some "pro-sumers" just fine. Pogue does a brilliant job of proving this by twice testing the public's ability to appreciate the megapixel difference in the final product: a printed photo, enlarged. Writes Pogue (registration may be required):
But one myth is so deeply ingrained, millions of people waste money on it every year. I’m referring, of course, to the Megapixel Myth.
It goes like this: "The more megapixels a camera has, the better the pictures."...
...I created three versions of the same photograph, showing a cute baby with spiky hair in a rowboat. One was a 5-megapixel shot, one was 8 megapixels and one was 13 [and had them blow up to poster-sized 16x24 inch prints]....
...[From a crowd of people that took the test] only one person correctly identified which were the low-, medium-, and high-resolution prints. Everybody else either guessed wrong or gave up, conceding that there was absolutely no difference.
Some of Pogue's readers challenged him. But the results were hardly different when he repeated the test with a slightly different methodology (also challenged). Although Pogue recommends the 5-7 megapixel range, he points out that he has seen photos from 4 megapixel cameras that look sharper than ones taken with an 8 megapixel camera.
So, how can this be? Well it really comes down what you do with all those pixels. Pogue made an incredibly important point when he blew his pictures up into poster-sized prints. For example, my 6 megapixel Nikon D70 can take a 3008x2000 pixel photograph. When all those pixels are finally displayed somewhere (on paper or on a screen), the sharpness of the image will be determined by how those pixels were ultimately treated. For example, if I try to print it on a poster that covers the entire wall of my son's bedroom (and I have a printer that can do that), the enlargement process will compensate by adjusting each pixel (and adding others) to a point that the image will look blurry or pixelated.
What Pogue proved in his test is that even if you enlarge a 5 megapixel photo to fit into a 16x24 inch poster, it looks just as good as if you enlarge an 8 or 13 megapixel image even though the latter image has more raw pixels to work with (in other words, fewer pixels to stretch, if stretching is even needed to fill the 16x24 frame).
Now comes the question of what most people do with the pictures they take. Hardly anybody I know ever enlarges their photos to something bigger than the biggest page size their printer can handle. In fact, most people do one or two of the following with the photos they keep: (1) print a 4x6, (2) print a 5x7, and/or (3) save it for viewing on the computer (from their hard drive, a CD/DVD, or a photosharing service on the Web). Believe it or not, in ALL of these cases (especially the digital viewing case), for most people, 3 megapixels will do. In fact, most of the pictures I take (many of which I load onto ZDNet) are taken at my D70's 3.3 megapixel setting and some are even taken at the 1.5 megapixel setting!
Consider this. If the most you think you'll ever do with your digital photos is look at them on your computer or share them on the Web, 99 percent of all digital photo viewing is done in a way that when the image is displayed, it's only taking up a fraction of the display. Then, take into consideration what resolutions most people have their entire display set to. Most people are set to 1024x768. At 786,482 total pixels, that's not even one megapixel. In other words, at 1024x768, you can't even view an entire 1.5 megapixel image without shrinking it to fit into your display. Go ahead, take a 1.5 megapixel image and try viewing it Firefox. In its full-screen mode, Firefox will automatically downsize it to make it fit in its window and give you a magnifying glass as the cursor so that you can return it to full size with a single left-click. When you're done clicking, you won't be able to see the entire image because it's to big to fit on the display.
OK, so you're runing your monitor at 1280x1024? That's 1.3 megapixels. (still way less than 3 or 4 megapixels). 1600 x 1280? That's 2 megapixels. Now take your typical photosharing service or any one of the photo applications that get used on the Windows or the Mac and 100 percent of the time, they default to shrinking images. For example, when you get to Flickr.com's image level page, Flickr actually shrinks (as do other services) images to be 500 pixels wide (dynamically auto-adjusting the height based on whatever the original image's aspect ratio was). A 500 pixel wide image is viewable in it's entirety at most of the commonly used resolutions (including the once popular 800x600 resolution). So, it was a good choice by Flickr. More importantly, to do this sort of shrinking, photosharing services must drop pixels. In other words, no sooner did you go out and buy that that 13 megapixel camera only to find out that your favorite photosharing service will throw out most of that data before it presents your pictures in the service's default viewing size.
The point is that as more and more people skip the idea of printing their pictures, the importance of having more than four or five megapixels is even further diminished.
So, when might you need more megapixels? Well, again, it comes back to what you're going to do with them. If you routinely crop your photos with a digital photo editor (eg: Adobe's Photoshop), then, both the pixel dimensions of the final product and how you're going to output the final product matters. If, for example, you started with 6 megapixels as I did when I took a picture of two bald eagles fighting in the air (see right, one eagle is talons-up) but, after cropping, ended up with a photo that's 168x132 pixels (the dimensions of the photo to the right), there isn't a whole lot more that can be done with this image. If I tried the Pogue test and stretched all 22,176 pixels of this image to fill a 16x24 inch poster, you probably wouldn't even recognize the subjects in the photos as birds (they'd be so blurry). But, if I had more original pixels to work with, I probably would have gotten better results (provided none of the blurriness is attributed to bad optics, the wrong aperature/ISO/shutter speed settings, or my inability to hold the camera still).
Here, for kicks, is what the same image looks like at 470 pixels wide (width in inches on your monitor will depend on your display's dot pitch, another factor in sharpness):
To expand what was once a 168 pixel-wide image into a 470 pixel-wide image without having a bunch of white space between the original pixels, Photoshop had to reach into its bag of tricks to synthesize new pixels and insert them into the image. It does this by inspecting the original image and making decisions based on what it sees. But, if I had more megapixels to work with from the getgo, such pixel fabrication probably woudn't have been necessary (or at least not as much).
Anyway, given how pixelated the above enlargement is, you can imagine how much worse it would look if we attempted to enlarge it to something poster-sized. Likewise, you can imagine the sort of tricks Photoshop would have ending up doing if all I gave it to work with was 1.5 megapixels. The above enlargement would have been far worse. That's why, when I saw this arial dog fight taking place and took note of how far away the eagles were (approximately 250 yards), the first thing I did was moan about not having my 500mm zoom lens with me. The second thing I did was crank the D70 up to its 6 megapixel setting (a setting I almost never use). I knew in advance that this is a picture that'd I'd be looking to crop because this is what I saw in the D70's viewfinder:
Now you know why a 500mm zoom lens would have come in handy. Short of having a good zoom lens, more megapixels for post-photo editing is probably the next best bet in terms of extracting good close-ups out of a far-away shot. Which brings me to my next point.
Two other features that you may notice when comparing digital cameras are their optical and digital zooms. With optical zoom, your lens is doing all the work and the ability to do good zooming will depend on the lens' optics. But with digital zoom, your camera is literally cropping what it sees before you take the photograph, and then its resizing that cropped image to match the whatever pixel dimensions your camera is currently set to take its pictures at. Much the same way Photoshop did some tricks to turn my 168 pixel-wide cropped photo into a 470 pixel-wide image, a camera's digital zoom does the same thing which is why its results will sometimes be on par with what happens after un-digitally zoomed photos are cropped and enlarged with a photo editor.
So, the basic question is what's going to happen with all the pixels? If the most that's going to happen to them is that they're going to be viewed on a computer or on a 5x7 print, 4 megapixels (heck, even 3) is actually more than enough. If you're going to be cropping and enlarging photos, then it might make sense to go with more megapixels. How many more? It's hard to judge. Only you know the sorts of pictures you think you'll be taking and to what extent you might be cropping and/or enlarging them.
Finally, as Pogue pointed out in his piece, there are so many other things that contribute to final photo quality that it might make sense to devote your time and/or the money you save (by going with fewer megapixels) on those. For example, the built-in flash on many cameras invariably does a fair job of lighting your subject, but not the sort of job that a high quality flash can do. With Nikon's wireless flash technology and its SpeedFlash external flashes, I've seen what otherwise might be a mediocre photo turned into a gorgeous shot simply because of the better lighting.
Update: I posted a part II to this blog entry. See The exception to the digital cam scam.