Thursday, October 30, 2008

Pixels

I was listening to a medical lecture and the doctor mentioned a study which showed that women had statitically fewer hot flushes on a particular medication but that it didn't make the women exactly pleased about the difference - the change? they had 11 hot flashes a day instead of 12 - but hey, it was a statistically significant difference.

This reminds me of some of the pixel wars and differences in number of pixels from one camera to the next. Anything less than a 25% difference and you probably can't even see it in tests and in real world shooting it's totally irrelevant - eg. the difference between 10 and 12 megapixel cameras, or for that matter, between 12 and 15 megapixels. Of course we know that to double print linear size (ie. 16X20 instead of 8X10 requires 4 times as many pixels for the same nose on print resolution (you are dealing with both length and width).

DPReview has just released their review of the 15 megapixel 50D, saying that in their opinion, the 50D is not sig. better than the 40D and in some ways not as good. Now, I couldn't quickly tell what lenses they used but it was clear from the examples that whatever lens they used, it wasn't good enough to take advantage of that many pixels. Additional sharpening might have made things look better but would not of course have added any more information to the images.

This finding, if it is replicated by other reviewers, will have profound influence on the future of the industry - specifically because it suggests that cameras which extend the pixel density of the 50D to full frame will be pretty much at the limit of useable pixel density. Frankly high resolution lenses which are only high when used at 5.6 and wider are not much help to a landscape shooter (or industrial or architectural).

It will be interesting to see how this plays out. It would seem that the limiting factor is diffraction but if that's the case, why is it that you can take the tiny sensor of a Canon G10 with 14.7 megapixels and produce decent images yet the far larger sensor of the 50D struggles - odd that. But how else can you explain that 60 megapixel medium format sensors can show improvements over the previous 39 MP backs.

I'd like to think that somewhere here is a loophole which will allow ever higher resolution ina 35 mm. format camera. Perhaps if the resolution is high enough we won't need fuzzy filters to avoid moire patterns but a recent interview with the head of Canon development suggested they aren't going to get rid of them any time soon. The assumption seems to be that medium format users are willing to take the time to shoot a reference colour card, to use software to fix moire etc. but that 35 mm. users aren't. There may well be some truth to that but for us landscape shooters it sure would be nice to have the option. Then there are those of us who'd seriously consider buying a full frame high pixel count camera that had no colour filters at all, ie. a full frame no bayer black and white sensor of 21 megapixels. Now that would be cool!

1 comment:

ajc666 said...

My understanding is that Leica abandoned the anti-aliasing filter in the M8. Unfortunately, they also removed the IR-removal filter as well. And the M8, once the IR-removal filter is attached to the lens, is supposed to produce exceptional images. I think the strength of the anti-aliasing filter may be one of the things that distinguishes cameras with the same sensor (e.g. a Sony CCD used in Sony, Pentax, and Nikon models).