Thread Subject: merge fluorescent images

hello,

i have a problem in merging 2 fluorescent images(red and
green images).
I have already analyzed those histogram in each layer.
All those significant layer in red image is in Red and for
green is in Green layer.

is it right if i combine those images by merge:
- Red layer from Red layer in red image
- Green layer from Green layer in green image
- blue--> assign zeros()

will there be a false colored?

thank you :)

~need reply soon

In article <g696sl$6ie$1@fred.mathworks.com>, mea <flymeafly@gmail.com> wrote:

>i have a problem in merging 2 fluorescent images(red and
>green images).
>I have already analyzed those histogram in each layer.
>All those significant layer in red image is in Red and for
>green is in Green layer.

>is it right if i combine those images by merge:
>- Red layer from Red layer in red image
>- Green layer from Green layer in green image
>- blue--> assign zeros()

>will there be a false colored?

Yes, of course. Any pixel that is active in both the red and green
layers will appear magenta (or is it cyan ? I can never keep track
of the way Red and Green mix in RGB.)


--
  "They called it golf because all the other four letter words
  were taken." -- Walter Hagen

The question is whether you want to keep the red plane in a one dimension of the array, red plane in the another dimension and blue in another dimension in an 3-D array. Or whether you want to keep the red plane in one one half of an array and green one in the another half of an array?

Yumnam Kirani Singh <kirani.singh@gmail.com> wrote in
message
<12028920.1216889432822.JavaMail.jakarta@nitrogen.mathforum.org>...
> The question is whether you want to keep the red plane in
a one dimension of the array, red plane in the another
dimension and blue in another dimension in an 3-D array. Or
whether you want to keep the red plane in one one half of an
array and green one in the another half of an array?


yes, it's true, there will be false color of course. thank
mr roberson :)


to yumnam, i want to do the former one.
ok, first i declare:
Green image: Rg Gg Bg
Red image: Rr Gr Br

merge image: Rr + Gg + Zeros()

is it right? because i have already analyzed those
histogram, and the most significant layer (which is most
information there) is in Rr and Gg, range are 0-255, and
distributes all area (dark-light area).

Other layers are not significant because those histogram
range are between 0-10 (dark area).

is it right if i merge with that way? can i analyzed it
based on those histogram?

thank u, sorry my english is not quite good :)

"mea " <flymeafly@gmail.com> wrote in message <g69hrf$kv9
$1@fred.mathworks.com>...
> Yumnam Kirani Singh <kirani.singh@gmail.com> wrote in
> message
>
<12028920.1216889432822.JavaMail.jakarta@nitrogen.mathforum.
org>...
> > The question is whether you want to keep the red plane
in
> a one dimension of the array, red plane in the another
> dimension and blue in another dimension in an 3-D array.
Or
> whether you want to keep the red plane in one one half of
an
> array and green one in the another half of an array?
>
>
> yes, it's true, there will be false color of course. thank
> mr roberson :)
>
>
> to yumnam, i want to do the former one.
> ok, first i declare:
> Green image: Rg Gg Bg
> Red image: Rr Gr Br
>
> merge image: Rr + Gg + Zeros()
>
> is it right? because i have already analyzed those
> histogram, and the most significant layer (which is most
> information there) is in Rr and Gg, range are 0-255, and
> distributes all area (dark-light area).
>
> Other layers are not significant because those histogram
> range are between 0-10 (dark area).
>
> is it right if i merge with that way? can i analyzed it
> based on those histogram?
>
> thank u, sorry my english is not quite good :)
>
I am not sure whether this is relevant or not, but one
thing you can do with your merged Red, Green and empty
image is to do a Principal Component analysis on the data
in colour space. This will seperate out any information
that is common to your red and green planes, and give you a
two plane image which contains no 'false colour'. e.g. your
Yellow data will disappear. (You was almost right Walter;-)

Regards

Dave Robinson

> I am not sure whether this is relevant or not, but one
> thing you can do with your merged Red, Green and empty
> image is to do a Principal Component analysis on the data
> in colour space. This will seperate out any information
> that is common to your red and green planes, and give you a
> two plane image which contains no 'false colour'. e.g. your
> Yellow data will disappear. (You was almost right Walter;-)
>
> Regards
>
> Dave Robinson
>
>

cmiiw,, what u mean is: i implement pca on those layer:
Rr Gg
is it right?
thank u

"mea " <flymeafly@gmail.com> wrote in message
<g6bt8e$lqb$1@fred.mathworks.com>...
> > I am not sure whether this is relevant or not, but one
> > thing you can do with your merged Red, Green and empty
> > image is to do a Principal Component analysis on the
data
> > in colour space. This will seperate out any information
> > that is common to your red and green planes, and give
you a
> > two plane image which contains no 'false colour'. e.g.
your
> > Yellow data will disappear. (You was almost right
Walter;-)
> >
> > Regards
> >
> > Dave Robinson
> >
> >
>
> cmiiw,, what u mean is: i implement pca on those layer:
> Rr Gg
> is it right?
> thank u

Sorry I don't actually understand what you are asking above.

What I am suggesting MIGHT be useful, if you have a lot of
yellow pixels then if you consider Red, Green, Blue to be
an orthogonal set of colour axes, then a significant amount
of correlation exists within your Red and Green axes. A
better set of axes can be defined, which minimises this
correlation by effectively rotating your orthogonal axes to
a position where this correlation is minimised - this is
what colour space PCA will do for you.

When you recombine the three new axes to form a new colour
space, and try to display it, the pixels will still come
out Red, Green and (virtually no) Blue, because thats the
way your computer hardware is built, but you should now
find any region of correlation between the new Red and
Green axes will be minimised, and hence significantly less
yellow pixels.

Hope that helps

Dave Robinson

On Jul 24, 2:19=A0am, "mea " <flymea...@gmail.com> wrote:
> hello,
>
> i have a problem in merging 2 fluorescent images(red and
> green images).
> I have already analyzed those histogram in each layer.
> All those significant layer in red image is in Red and for
> green is in Green layer.
>
> is it right if i combine those images by merge:
> - Red layer from Red layer in red image
> - Green layer from Green layer in green image
> - blue--> assign zeros()
>
> will there be a false colored?
>
> thank you :)
>
> ~need reply soon

----------------------------------------------------------------
mea:
=46rom your description it appears that you want to average two color
images. You want to average the red channels of the two images and
average the green channels of the two images, and set the blue channel
of the result image to zero. This will give the appearance of a
"merged" image however unless the images are identical you will have
color and edge effects. If the images are really identical (as far as
being spatially colocated) and you are averaging frames to reduce
noise then this will work. If one image is slightly shifted,
magnified, or rotated from the other, then that is when you will see
effects of misalignment in your result image. If that is the case you
will need to align your images. Do a web search for "image
registration," "register images," or "align images" for tons of info.
Image registration is a big topic in image processing.
Regards,
ImageAnalyst