Originally Posted by
dblevitan
Noise is actually a function of much more than the pixel size. The quality of the electronics, the quality of the sensor, and even temperature (among many other factors) all contribute to noise. It is easier to have a higher signal to noise ratio with larger pixels because you can collect more light but its also very possible to have more pixels and lower noise. Astronomical CCDs routinely perform very long exposures (minutes long) and after fairly routine noise subtraction produce very clean images.
the dominant source of noise in a digital camera is photon counting statistics which is directly related to pixel size. other factors such as read noise, thermal noise, etc. are negligible, except in the darkest areas. see roger clark's excellent analysis:
http://www.clarkvision.com/imagedeta...mance.summary/
This is true, but he later published another article where he admits that this is not the best way to do it. In that article, he describes how one of his readers worked with him to perform another test. In that test, they took one camera and shot the same object multiples times such that when cropped around only that object the same picture was presented but with 3 different megapixel counts. Only 3/50 people correctly figured out which picture had which resolution.
cropping and enlarging is not the same as using actual cameras because now you've introduced different sensor sizes. nevertheless, for prints 8x10 or smaller, it will be virtually impossible to discern a difference between 6 and 10 megapixels. however, for larger prints or if one crops, it is likely to be noticeable.