How to copy colormap from an existing image

I have an image that came from an imaging software that already comes complete with its own color bar. Below is an example so you know what I mean. I could use any image software to grab some RGB values from various points within the color bar. But then I'm not sure how to proceed. I can't figure out how to use the Colormap Editor at all... Like how do you put new colors into it?? I've browsed through some other questions that discuss how make color maps through commands, but the ones I've seen are for very specific discrete colors that the user knows. Here, I want to replicate the entire color bar--so probably 256 colors. How many colors do I need to sample to ger a reasonably accurate copy of this? And how do I interpolate them?

(The custom colormap doesn't have to match this exactly, just good enough so two images look like they're using the same colormap without any obvious color differences.)

Answer 1

[David Goodmanson]

HI Alexei

I copied the image you provided as a jpeg, used datatips to find coordinate points at the top and bottom of the colorbar and ran the following code. It shows the profile of the color map, with color intensity on the vertical and colormap row index on the horizontal. There are 210 rows, although the number of rows does not make too much difference since it is the scaled distance from 1 to rowmax = 210 that matters.

MODIFIED to incorporate Walter's comment below

micro_m = imread('micro_m.jpg');
image(micro_m)
% datatips show coordinates (20,16) and (20,225)
cmap = squeeze(micro_m(225:-1:16,20,:));
cmap = double(cmap)/255;              % 0 <= values <=1 for colormap 
n = size(cmap,1);
ind = 1:n;
figure(2)
plot(ind,cmap(:,1),'r',ind,cmap(:,2),'g',ind,cmap(:,3),'b');grid on
grid on

The colormap somewhat resembles 'jet' but is not the same.

[Walter Roberson]

Notice that David's cmap is just an array with 3 columns. And that is all a colormap is, an array with 3 columns. However, colormaps should contain values in the range 0 to 1, so take the cmap that David computes above, which will be uint8, and calculate

cmap = double(cmap)./255;

and the result will be something you can use with colormap()

[Alexei M]

So I looked at it like this, which I think is what you're referring to when you say you "used datatips":

I'm not seeing how you get coordinates of (20,16) to (20,225). The maximum x-coordinate that is still on the color bar is 16. In the y-direction, it goes from 9 to 223.

But in any case, that's an easy fix. However, I'm not quite following the line with the squeeze() function. Can you explain that a bit more?

[David Goodmanson]

Hi Alexei, yes that's what I was doing with datatips, but the coordinates can certainly vary depending on the way the image is obtained. I got your image with the Windows snipping tool so that could have something to do with why your colorbar coordinates are different. What matters is to just find a vertical line that runs down the middle of the colorbar. Not the most sophisticated way to go about it I suppose, but it got the job done. In your case x = 10 looks good.

To describe

cmap = squeeze(micro_m(225:-1:16,20,:));

I will use the dimensions of the jpg I had, which was 246x117x3. x = 20 was a good datatips column index in my case. Then

micro_m(225:-1:16,20,:)

pulls out just that one column, leaving an array that would have been 246x1x3. Except some selection on the first index cuts it down to 210x1x3. The colormap itself is going to be 210x3, so squeeze just gets rid of the second index 1, which only has one value.

[Alexei M]

Oh I see. But for some reason I still don't get this part:

225:-1:16

Why is there a -1?? And how can you have 2x of ":"? If you want to extract that one vertical line from the image array, I thought the argument would be:

(16:225,20,:)

Because you want all the array values ranging from pixel 16 down to pixel 225, don't you? The column value stays fixed at 20, I get that. And the ":" in the 3d dimension to take all the values--I get that part, too.

[Image Analyst]

That's standard MATLAB syntax. It's basically startingValue : increment : endingValue. So it's [225, 224, 223, ..... 19,18,17,16]. You could do it like you did but then the first element of the colormap would be the color at the top of the bar, and this is not the convention in MATLAB. The first row in the N-by-3 colormap is the RGB color of the bottom of the colorbar, not the top of it.

[Alexei M]

Ohh... the increment! I forgot all about that part of the syntax.

Ok now it's clear, thanks! (Sorry about the silly question. Should've known that.)

Answer 2

[Image Analyst]

See my attached demos where I grab the colorbar from an image and create an colormap from it. Then I invert the colormap to turn the pseudocolored image back into the gray scale/indexed image.