On Fri, 30 Jul 2010 11:03:33 +0200, Ofnuts
wrote:

On 30/07/2010 10:36, bobwilliams wrote:
bobwilliams wrote:
Let's assume I have a 10MP camera
My sensor is say, 3650 X 2740 pixels.
But say I want to create an image at 1825 x 1370 pixels.
How does the camera actually reduce the 5.0MPs to 2.5MPs
Does it choose groups of 4 pixels and somehow average them out to
groups of 1 pixel each?
How does this process differ from compressing the 10MP image by a
factor of 4.
I know that in one case the image SIZE is reduced (as well as the file
size) whereas in the other case, the image SIZE remains the same but
the file size is reduced.
How exactly does each process affect the appearance of say an 8x10 print.
Bob Williams

OOPS!
I meant to say, How does the camera actually reduce the 10MPs to 2.5MPs?
Bob

When using a smaller pixel count, what a given camera does exactly is
hardly documented. Most processing would be built around these
algorithms (in decreasing order or quality and computing power needs):

http://en.wikipedia.org/wiki/Lanczos_resampling
http://en.wikipedia.org/wiki/Bicubic_interpolation
http://en.wikipedia.org/wiki/Bilinear_interpolation
http://en.wikipedia.org/wiki/Nearest..._interpolation

PC software normally uses the first two. Some pre-blurring, and
post-sharpening can be used. This would also be combined with the
demosaicing algorithm.

There are dozens more resampling algorithms than that available in good
computer software. Like Bell, Spline, Pyramid, Mitchell, Vector, Triangle,
Hermite, Catmull-Rom, and all their various incarnations. Bicubic being the
most often used in-camera algorithm due to its speed (vs. poorer
detail-retaining performance) that is required for fast in-camera
processing. I know of NO camera that would ever use the processing
intensive Lanczos algorithms. That's just wishful dreaming on your part.
Nor do I know of any that would use the rudimentary Bilinear or Nearest
Neighbor. Those are just too crude for any in-camera processing.

Please only offer advice and opinions on things that you actually know
about. It's so tedious having to continually correct you lousy ****-head
pretend-photographer role-playing trolls.