how to display a grayscale image in a monochrome color (other than white)

I have a grayscale image consisting of a matrix of 512x512 int16 numbers. If I run

figure;imshow(imagedata,[])

I can see my contrast adjusted image. Now all I want to do is to display this image by replacing white with green. I do not want to change the datatype to double or do something that would mess with the original data.

My first thought was to just look at the colormap and ideally that should look like

[0 0 0]
[1 1 1]
....
[255 255 255]

I thought I could just change that to be

[0 0 0]
[0 1 0]
.....
[0 255 0]

but when I take a look at the colormap of this grayscale image,

map = colormap;

the colormap has non-zero numbers (which are not equal to each other) in each of the r,g and b columns and so I don't understand this. The grayscale colormap is supposed to have the same numbers in the r,g and b columns.

All I want to do is to tell matlab that the highest number maps to green instead of white. What is the easiest way of doing this?

Answer 1

[srt10]

After much searching, I found the answer to my question. There is a function called

cmap2gray

that was introduced in R2020b (and I was using R2020a so I didn't find it right away).

The map which is output when you use the colormap function on a grayscale image is an "RGB colormap" and can be converted to a "grayscale colormap" by running

newmap = cmap2gray(map)

The 3 columns of newmap are identical (which is what you would expect). Now, I can easily convert my grayscale image which is originally black to white, to black to green by zeroing out the R and B columns in the newmap variable.

I can then apply this corrected "grayscale colormap" to my grayscale image to get it to display in green.

[John D'Errico]

Sad. You clearly did not understand the answer Voss gave, which DID answer your question.

[srt10]

@John D'Errico, can you please explain what I missed?

[John D'Errico]

Voss spent a great deal of effort in explaining it, over and over again. Then you decided to post an answer that effectively does something very similar, while claiming that Voss did not even make an effort to answer your question.

[DGM]

Using cmap2gray() is not appropriate here, and the results it returns may not be linear or possibly not even monotonic. We'll get to that.

Voss is correct in that the colormap() function does not return any information about an image, but I disagree with the use in the example. It does one of two things. It can either be used to specify the colormap for a given axes or figure (e.g. the current axes), or it can be used to retrieve the current colormap from a figure or axes.

When you use colormap() to set a colormap without specifying a target axes or figure, it sets the colormap of the current figure, which in turn also sets the colormap of the descendant axes. When you call colormap() with an output argument, you're just getting whatever the current colormap is. As before, if you do so with no input argument, the default target is the current figure.

So what is the current colormap used by imshow()? Well, if you call imshow() with a single-channel image with the following syntax:

imshow(grayimage) % default scaling or
imshow(grayimage,[]) % scaled to data extrema

then imshow() will set the axes colormap to gray(256), and CDataMapping is set to 'scaled'. This is an example of scaled colormapping.

If you call imshow() like so instead:

imshow(indexedimage,colortable) % an indexed-color image

then imshow will set the axes colormap to the specified colortable, and CDataMapping to 'direct'. This is an example of direct colormapping.

In both of these instances, we're talking about the current axes colormap. Remember, the thing that you get when you call colormap() like this

CT = colormap(); % get the colormap for GCF

is the current figure colormap.

So what is the current figure colormap? That's hard to say. It could be any number of things if the figure has been used for other things prior. Chances are it's still the default colormap, which is parula(256).

What does cmap2gray() do? This tool does the same BT601 luma transformation that rgb2gray() and im2gray() do; the only difference is that it's tailored to accept inputs as Mx3 unit-scale colormaps, and it expands the output so that it stays Mx3.

What happens if you process one of the typical colormaps with this tool? Well, since none of the standard colormaps are linear and full-scale in BT601 luma (except gray()), the result will not be a linear grayscale ramp.

% a single-channel grayscale image
% a full-scale sweep from black-white
graypict = repmat(linspace(0,1,200),[100 1]);

% a colormap
CT = parula(256);

% display the image, rendered with the specified map
imshow(graypict)
colormap(CT)

% convert the colormap to gray
CT = cmap2gray(CT);

% display the image, rendered with the specified map
subplot(2,1,1)
imshow(graypict)
colormap(CT)

% plot the luma
subplot(2,1,2)
plot(CT(:,1))
ylim([0 1])

As you can see, the resulting map is not full-scale, linear, or monotonic. It's not really useful for much of anything. So why does cmap2gray() even exist? The example given in the documentation is pretty much it. It's useful for reducing the color tables associated with indexed-color images. That's not your application.

So if you wanted a gray colormap to modify, how would you get it? You'd use gray().

mycolormap = gray(256); % a gray colormap
mycolormap(:,[1 3]) = 0; % make it green

% display the image with the custom map
imshow(graypict,[]) % scaled to data extrema
colormap(mycolormap) % use the new map

Although creating colormaps by editing gray() is limited and cumbersome. A more useful and convenient approach would be something like:

CT0 = [0 0 0; 0 1 0.5]; % pick any endpoint colors you want
N = 256; % specify the colortable length

% generate a custom colormap
CT = interp1([0 1],CT0,linspace(0,1,N));

Answer 2

[Voss]

Like this?

imagedata = randi([0,65535],512,512,'int16'); % random 512-by-512 int16 matrix

figure; imshow(imagedata,[])

cmap = colormap();
cmap(:,[1 3]) = 0;
colormap(cmap)

[srt10]

I don't think that's actually doing the right thing. For example, please look into your cmap variable and tell me why the R, G and B columns are different numbers instead of the same numbers?

[Voss]

imagedata = randi([0,65535],512,512,'int16'); % random 512-by-512 int16 matrix

figure; imshow(imagedata,[])

cmap = colormap();
cmap(:,[1 3]) = 0

cmap = 256×3

0 0.1504 0 0 0.1534 0 0 0.1569 0 0 0.1607 0 0 0.1648 0 0 0.1689 0 0 0.1732 0 0 0.1773 0 0 0.1814 0 0 0.1854 0

colormap(cmap)

The R, G, and B columns are different because the R and B have been zeroed-out, in order to make shades of green. If R, G, and B were all the same, then you'd have shades of grey.

[srt10]

I don't think the function colormap is designed to take grayscale images as input. So therefore, although it returns a map that looks legit, performing these kinds of manipulations is going to have unintended consequences.

[srt10]

@Voss please look at the cmap variable before you zero the numbers out. Since it is a grayscale image, the three columns should have the same numbers. But they aren't. I don't know how else to explain it to you. You need to first see the data to make sure the zeroing out operation actually is doing what you want it to do. Please read my question again carefully

[Voss]

A colormap is an N-by-3 matrix of numbers. The default colormap is not greyscale, hence the columns are different. You are confusing the colormap with the image data.

[Voss]

"I don't think the function colormap is designed to take grayscale images as input."

The colormap function is not designed to take any images as input. That's why you get this error when you run the code in your question:

imagedata = randi([0,65535],512,512,'int16'); % random 512-by-512 int16 matrix

map = colormap(imagedata);

Error using colormap
Colormap must have 3 columns: [R,G,B].

[srt10]

@Voss sorry about my wording. I don't think you are understanding the point.

Take your variable

imagedata = randi([0,65535],512,512,'int16'); % random 512-by-512 int16 matrix

Since it has int16 data, it is a grayscale image.

now plot it and calculate its colormap

figure; imshow(imagedata,[])
cmap = colormap();

When you plot it, you will see a greyscale image. Now the cmap variable, since it is a greyscale colormap, should have R,G and B values the same. But it doesn't. Why?

[srt10]

@Voss I thought it was pretty clear from the original question that I know I can try to zero out the R and B columns in the colormap to make the image look green. The question asked is subtle because the colormap function doesn't behave in an intuitive way for grayscale images. I think you are missing that point. I have edited my question to correct the matlab syntax

[Voss]

"calculate its colormap"

The output of the colormap function has nothing to do with the image data.

[srt10]

I give up. You are being intentionally obtuse

[Voss]

You are misunderstanding what the colormap function does. Read the documentation. Here's a link: https://www.mathworks.com/help/matlab/ref/colormap.html

[srt10]

I don't think I am misunderstanding anything. I think you are being unhelpful by not even attempting to understand the question

[Voss]

I think you are wrong about that, too.

[srt10]

thanks for your opinion. It was very helpful. Just like your contribution to this question