Indeed, this kind of information is expressed by M1 and M2 (what the 
community named the "Manders' coefficients"). However, be aware that you 
should be very careful with your conclusions from your measurements. Like all 
co-localization coefficients these coefficients are very sensitive for off-set 
and cross-talk. A high offset gives a high background and background 
correlates and may co-localize with backgound and this will strongly bias your 
coefficient (especially in 3D data-sets since there are lot of background pixels 
in 3D-images). Cross-talk is undistinguishable from co-localisation and should 
therefore be reduced during scanning (sequential scanning) of removed by 
post-processing. So you should allways measure your cross-talk in control 
experiments (single colour labelling) with the same microscope settings (laser 
power, pinhole, PMT voltage, etc) as your real experiment. Another point of 
attention is the threshold that has to be set in the Manders' coefficients. This 
may be a very critical setting, so be careful and do your control experiments. 
The numbers that come out should be compared with these control 
experiments. And this counts for all colocalisation coefficients. You should 
compare the numbers with the numbers of control experiments. There are two 
kind of experiments that you may do. 1) Biological control experiments: e.g. Do 
another pair of stainings of which you know that they should colocalize and do 
a pair of stainings of which you know that they cannot colocalize at all. Then 
measure your relative "zero" and "one". 2) Image processing experiment (only 
possible to measure your "zero"). Take two dual-colour images of look-alike 
cells. Take the red component of the first and measure the colocalisation with 
the green component of the second image. Repeat this for some more 
combinations of images and see what the colocalisation coefficents are of two 
uncorrelated patterns. It is clear that you apply the same post-processing 
(cross-talk reduction, noise reduction, off-set reduction, deconvolution, etc) 
and colocalisation measurement (thresholds, etc).
So, my take home message is: It is easy to get numbers from an image. The 
interpretation of these numbers needs a lot of attention.

Good luck and succes with your research, 
Erik Manders