Having founded photo.net back in 1993, it has long surprised me that there is no “photo.net for camera phones”. DxO Mark has begun to put up some objective tests that are interesting, though they don’t capture the practicalities of using camera phones. The tests show that the Nokia PureView camera, which averages 41 megapixels of crummy noisy data, crushes the competition in low light (i.e., the indoor conditions where most people want to take pictures). The Samsung Galaxy S III is a distant second in the noise department. Yet worse (though not by much) in low light are the latest Apple products, iPhone 5, iPhone 4S, and iPad 3, all of which offer similar performance.
(Recent tests on the same site show that Canon continues to lag in sensors. The Powershot G15 is absolutely crushed by the Sony RX100, admittedly $200 more expensive. Canon either needs to improve its technology or buy DxO and shut down their web site!)
I thought the large-sensor Canon G1X was the RX100 competitor?
Demetri: I think the G1X is huge compared to the RX100 and G15. So it is competitive in terms of price with the RX100 but not competitive in terms of being a pocketable camera. The Sony is about half the weight of the Canon. With the APS-C sensor, the G1X gets a low-light ISO rating of 644 and an overall DxOMark score of 60. The RX100 is a half-stop worse in low light (ISO 390) but gets an overall score of 66 due to its superior dynamic range (a strong point for Sony/Nikon; a weak area for Canon lately).
The Galaxy S III has lower noise because it runs very aggressive noise and sharpening on the images.
Mr. Greenspun
Have you reviewed the testing procedure outlined by DXO?
I’m curious how you feel about normalizing all images to 8mp before doing their measurements? Wouldn’t this disadvantage large megapixel phones and favor the iphone etc?
Also, they seem really obsessed with accutance, but if my understanding is correct that’s just the appearance of sharpness not a measure of resolution?
For myself I find those two aspects of their reviews as kind of odd as I’ve never seen either technique (“normalizing” and accutance) used by camera review web sites.
Thanks!
Joe: I haven’t looked too carefully at their protocol. It seems to be a PDF trapped inside a tiny browser window. If this is the best that they can do for Web publishing perhaps it is time to be skeptical about their camera tests! Then of course I’m using this on Windows 8 so when I downloaded the file and opened it this window into which I am typing a reply disappeared in favor of a mostly blank screen with the PDF in the middle…
I think it is reasonable to bash everything down to 8 MP. There is no way to get more than 8 MP of high quality information out of current devices (I think that the point of Nokia’s 41 MP is not to compete with the Nikon D800 but rather to deliver a practical benefit in moderate resolution images). With a linear measurement of 4000 pixels, for example, it is possible to get traditional photo quality in an enlargement 20 inches wide (200 pixels per inch). It is hard to believe that the tiny sensors in these cameras are going to justify more enlargement than that.
Given that they are norming everything to 8 MP, concentrating on acutance does not seem like a bad idea. They have set up a situation where all of the tested phones will have the same 8 MP resolution (even if they actually don’t!). So what is left other than to look at comparative acutance?
I’m not sure that their overall weighting/scoring makes sense. Almost all smartphone cameras produce pretty good images in bright light. The weak points would seem to be (1) dynamic range (highlight and shadow detail), and (2) image quality in low indoor light. Why not a system that gave most of the weight to those issues?