BitDepth 600 - October 30
07/01/09 19:25 Filed in: BitDepth - October 2007
David Pogue considers just how many megapixels you really need in a digital camera.
Pogue tackles the Megapixel Myth
David Pogue talks to members of the audience after his keynote presentation on the Megapixel Myth at PhotoPlus Expo 2007. Your Mac Life Show host Shawn King is seen in the background. Photo by Mark Lyndersay.
For the second keynote session at PhotoPlus Expo, the show's organisers took a chancy and ultimately quite agreeable leap into the populist, engaging New York Times correspondent David Pogue to speak on one of his more popular blog (and later column) topics, "The Megapixel Myth."
Pogue isn't a geek writer rousted out from behind a musty keyboard to expound on a personal bête noire. He's a journalist with a long background in the Mac industry, a former Broadway arranger and conductor and a comfortable public speaker, so the event was quite the show.
Animated, passionate to the point of fevered, Pogue proceeded to engage his pet theory, which posits that "for the non-professional photographer, five to eight megapixels is more than enough."
The whole thing began with a blog entry on the New York Times website which drew 500 comments and 700 e-mails, many of which, apparently, dismissed the author as a doofus. The reason? His test of the theory, three images of a pretty young child that he down-sampled to provide his test material and reproduced as large prints offered for public comment, was technically flawed.
This allowed Pogue to take the position of the embattled champion of the obvious, which he clearly relishes, given his theatrical gestures and head tossing as he engaged the audience.
So Pogue repeated the test using images provided by Professional Photographer's Ellis Veneer, who shot a new image (emphasising technical, not cute) by using first the full frame of a Canon 1DS and then backing off to use less and less of the frame. This gave Pogue three printable files that used the same camera sensor and optics and, as it turns out, essentially the same result, minus the cute baby.
At the back of the lecture hall, the three images were printed large and mounted for viewing. As with the first test, most people were reduced to guessing which was which.
Now setting aside (reluctantly) Pogue's amusing show closer, rewrites of popular songs that lampoon the tech industry (e.g. The Bill Gates song, or "I write the code"), there's a crucial issue at stake here as camera manufacturers cram more and more light sensitive photosites into the tiny space available for a sensor in most point and shoot cameras.
The dirty secret of modern consumer digital cameras capturing more than eight megapixels is that they tend to clog highlights and generate unacceptable noise, a scattering of colourful pixels in areas that should be solid black.
And there's another angle on this that Pogue didn't touch at all, and that's the fact that digital camera's don't actually capture megapixels, that measure was just a way of reducing the unwieldy size of modern captures to something that you read without immediately going cross-eyed.
So five megapixels actually refers to a file size of 5,000,000 pixels, but a camera's sensor will divide its available photosites according to manufacturer strategy for capturing various percentages red, green and blue, with each sensor capturing only one of the three colours. The camera's internal processors than take that capture information and use prediction analysis (tech for "guess") at what the image is supposed to look like.
Just to guarantee that technology remains impenetrable, even if you managed to go along with the technology used in most cameras, one manufacturer, Foveon, uses a chip that captures all three colours at each photosite, but makes chips which use fewer active photosites per square millimeter and require extrapolation on their length and width in processing to create competitive file sizes.
All these variables are why professionals like to work with RAW files, which allow at least some control over the way that captures are interpreted to create a final working image file.
For everyone else, the take-away seems to be that until some new corner in technology is turned, packing more than eight megapixels worth of photosites onto the surface area of a typical consumer digital camera sends image quality into a spiral of diminishing returns.
Links
David Pogue's original blog posting is here...
Read Mike Chaney's more technical overview of "The Megapixel Masquerade" here...
David Pogue talks to members of the audience after his keynote presentation on the Megapixel Myth at PhotoPlus Expo 2007. Your Mac Life Show host Shawn King is seen in the background. Photo by Mark Lyndersay.
For the second keynote session at PhotoPlus Expo, the show's organisers took a chancy and ultimately quite agreeable leap into the populist, engaging New York Times correspondent David Pogue to speak on one of his more popular blog (and later column) topics, "The Megapixel Myth."
Pogue isn't a geek writer rousted out from behind a musty keyboard to expound on a personal bête noire. He's a journalist with a long background in the Mac industry, a former Broadway arranger and conductor and a comfortable public speaker, so the event was quite the show.
Animated, passionate to the point of fevered, Pogue proceeded to engage his pet theory, which posits that "for the non-professional photographer, five to eight megapixels is more than enough."
The whole thing began with a blog entry on the New York Times website which drew 500 comments and 700 e-mails, many of which, apparently, dismissed the author as a doofus. The reason? His test of the theory, three images of a pretty young child that he down-sampled to provide his test material and reproduced as large prints offered for public comment, was technically flawed.
This allowed Pogue to take the position of the embattled champion of the obvious, which he clearly relishes, given his theatrical gestures and head tossing as he engaged the audience.
So Pogue repeated the test using images provided by Professional Photographer's Ellis Veneer, who shot a new image (emphasising technical, not cute) by using first the full frame of a Canon 1DS and then backing off to use less and less of the frame. This gave Pogue three printable files that used the same camera sensor and optics and, as it turns out, essentially the same result, minus the cute baby.
At the back of the lecture hall, the three images were printed large and mounted for viewing. As with the first test, most people were reduced to guessing which was which.
The dirty secret of modern consumer digital cameras capturing more than eight megapixels is that they tend to clog highlights and generate unacceptable noise, a scattering of colourful pixels in areas that should be solid black.
And there's another angle on this that Pogue didn't touch at all, and that's the fact that digital camera's don't actually capture megapixels, that measure was just a way of reducing the unwieldy size of modern captures to something that you read without immediately going cross-eyed.
So five megapixels actually refers to a file size of 5,000,000 pixels, but a camera's sensor will divide its available photosites according to manufacturer strategy for capturing various percentages red, green and blue, with each sensor capturing only one of the three colours. The camera's internal processors than take that capture information and use prediction analysis (tech for "guess") at what the image is supposed to look like.
Just to guarantee that technology remains impenetrable, even if you managed to go along with the technology used in most cameras, one manufacturer, Foveon, uses a chip that captures all three colours at each photosite, but makes chips which use fewer active photosites per square millimeter and require extrapolation on their length and width in processing to create competitive file sizes.
All these variables are why professionals like to work with RAW files, which allow at least some control over the way that captures are interpreted to create a final working image file.
For everyone else, the take-away seems to be that until some new corner in technology is turned, packing more than eight megapixels worth of photosites onto the surface area of a typical consumer digital camera sends image quality into a spiral of diminishing returns.
Links
David Pogue's original blog posting is here...
Read Mike Chaney's more technical overview of "The Megapixel Masquerade" here...
blog comments powered by Disqus
