Re: Merging of Images

Posted by Gabriel Landini on Jan 31, 2011; 12:32pm
URL: http://imagej.273.s1.nabble.com/Merging-of-Images-tp3685871p3685872.html

On Monday 31 Jan 2011 11:55:00 Phase GmbH wrote:
>  I am faced with the following problem:
> A microscopic sample is labelled with 2 fluorescent dyes, one emitting red
> the other yellow. 2 Images are taken by a black/white camera, one for the
> red label one, one for the yellow label. Every Image is colourized by a
> applying the corresponding LUT.
>
> My question is know: What would be the best procedure to merge both images?
> The final image should show both labels in their correct colours.

You can mix yellow and red but this will be a confusing orange, as the hues
are not that far away to produce a new distinct additive colou mix.

I would rather put the "red" greyscale image (i.e. without LUT) in the blue
channel and the yellow signal (which you caputured in greyscale as well) in
both the red and green channels.
This way ther "red" appears as "blue" and the "yellow" still as "yellow".
Cololocalisation appears as white. This can be seen correctly by the vast
majority of viewers, including all red-green colour blinds (e.g. protanopes,
deuteranopes, and anomalous prot- and deu- dichromats). Another advantage is
that they will perceive the colours with the correct names (unlike with the
green-magenta pairs that some use). Tritanopes however (a very small
proportion of dichromatic viewers) might struggle to see the differences.
There is no solution that fits all and preserves colour name perception.

Note that the "correct colours" is somewhat misleading as there is no
guarantee that the RGB LUT matches the emission spectrum of the dyes.
I gave a talk on this subject at the last Luxembourg meeting.

Regards

Gabriel