www.ukbutterflies.co.uk

This is obviously a valuable technique — NASA uses it routinely.

In brief, fewer but larger pixels result in less noise. So a 4 x 4 binning produces larger pixels (twice the linear, four times the area). Some newer cameras actually offer this facility, but most don’t.
A 12 mp (eg 4,000 x 3,000) image would thus be reduced to 3 mp (2,000 x 1,500) but this would not be the main reason for wanting to do it; less noisy images would be the aim.

I simply cannot find any software to do (say) 4 x 4 binning post processing. Surely, it must be exist? Help!

Jack

The open-source UFRAW software can do various types of interpolation on RAW image files. See http://ufraw.sourceforge.net/Guide.html for more info. I use UFRAW in conjunction with GIMP software as a powerful 'free' alternative to Photoshop.

Mike

Thanks for the link Mike, but it's far too complicated for me.

Jack

I would think that 4x4 binning would compress 16 pixels to one. Your numbers would appear to describe 2x2 binning. Or is there something non-intuitive in the nomenclature?

It is also important to remember that Bayer sensors have one red, one blue and two green pixels for each four pixels. So 2x2 binning reverses that and actual 4x4 really improves things.

I almost never save full resolution JPGs. First I generate tiffs out of RAW file and then resize to my standard size ... which happens to be quite close to half. It is not as good as binning directly from RAW with exact half, but it works in the same direction.

You are correct. I did mean 2 x 2.

Jack

The binning you describe is just a simple averaging when downsizing an image. Any image editing software will do something at least as good when downsizing.

For example, in Photoshop (or Photoshop Elements), when you resize an image, you are given the choice of resampling algorithm to use: Nearest neighbour, Bilinear, Bicubic, Bicubic sharper, or Bicubic smoother. I'm not sure, but bilinear might be the same as the binning / averaging you describe if you downscale by an integer factor. I'd trust bicubic resampling to give a better result than simple binning though.

Thanks for pointing me in the right direction. I do in fact use bicubic for resizing but didn't previously understand why that was best.

Wiki:
http://en.wikipedia.org/wiki/Bicubic_interpolation
suggests that bilinear interpolation equates more-or-less to the pixel binning that I was talking about and agrees that bicubic is usually better.

Jack

Would be great if the camera had the option to do this
My 7D has 18 mp if it could use say 6 or 9 m pixels at high ISO and discard the noisy ones that would be really handy !