For Bayer sensors, 6MP is 1.5M red, 1.5M blue and 3M green photosites.  The
interpolation will generally use more than 4 photosites to get one full colour
pixel (see comparison of techniques at http://www-ise.stanford.edu/~tingchen/).
If you consider the sensor to only be 1.5MP then you are throwing away a lot of
luminance resolution.
For Foveon sensors the Mpixels claims are inflated as they count photosites as
pixels, even though the photosites are layered on top of each other so do not
contribute to added resolution.

Caveman, have you tried comparing large (say 12") prints from the 5MP and the
3MP?
If they are still similar try upsampling them both to 8MP.  At this point you
should definitely see the advantage to have 5MP of data to start with.

Dan

Quoting "J. C. O'Connell" <[EMAIL PROTECTED]>:

> Isnt it a fact that all digital sensors are rated in Mpixels
> but that is counting each of the mono red, green and blue
> pixels and when they convert to full color the image is interpolated
> upward to achive the same "Mpixel" figure. i.e. since it takes
> 4 mono pixels to create one color pixel, a 6Mpixel rated sensor
> is really a 1.5 Mpixel color sensor but the marketing guys
> like to keep that a secret.
>
> JCO
>
>
> -----Original Message-----
> From: Caveman [mailto:[EMAIL PROTECTED]
> Sent: Sunday, August 22, 2004 1:48 PM
> To: [EMAIL PROTECTED]
> Subject: Re: sad stuff about stock photography and up-to-date technology
>
>
> Funny, this morning I was doing a similar experiment but in the opposite
>
> direction. I was curious if the 5MP from a 5MP digicam (Canon S60 in
> this case) are really holding 5MP of real information or if there are
> "invented" pixels there. So I took pics with the same subject at nominal
>
> 5MP and the other MP sizes supported by the camera (3MP and 2 MP, and I
> didn't bother with the 640x480). I resampled the 3 and 2 MP images to 5
> MP and compared with the genuine 5MP image. The 2 MP obviously lost some
>
> fine details, however the resized 3MP one was oh so similar to the 5MP
> one, except some JPG artifacts in a grass covered area. Except those
> artifacts, the fine details were virtually identical.

Reply via email to