how to add a custom colormap to 16 bit grayscale image without using imshow

I have a 16 bit image and want to create a colormap based on pixel intensities.

I am interested in a narrow band betwen 0 and 6000 pixel intensity, and want to have a colormap on the image for further analysis on the picture such as adding annotations etc.

The only way I found that works, is by using imshow. The problem is that I don't want to display the image, I want to do further analysis.

Is there a way to do what imshow does, without actually showing the image?

imshow(image, 'DisplayRange',[0 6000],'Colormap',customColormap);

Answer 1

[Image Analyst]

Try this. I made a 400 row by 600 column uint16 image with values from 1 to 6000. Then I took a range of values from 3000 to 3805 and pseudocolored them with the "jet" colormap. I used 3805 just to make it bigger so you can see it in the image, but feel free to use 3005 as long as you have enough pixels in that range to see a difference. Of course you could colorize outside that range with a different colormap if you want or even a colormap that's all black to blacken everything outside the desired range. Let me know if this meets your needs.

% Create sample data.
data = linspace(1, 6000, 400*600);
data = reshape(uint16(data), 400, 600);

% Show image in it's original gray scale.
subplot(2, 1, 1);
imshow(data, []);
colormap(gray(256));
colorbar;
title('6000 gray levels in gray scale', 'FontSize', 15);

% Show image where a specified range is in pseudocolor.
subplot(2, 1, 2);
cmap = gray(6000);
row1 = 3000; % Whatever.
row2 = 3805; % or 3005 or whatever you want.
% Make the jet colormap with the proper number of rows.
cmap2 = imresize(jet(row2-row1+1), [row2-row1+1, 3]);
% Paste that over the appropriate rows of the gray scale colormap.
cmap(row1:row2, :) = cmap2;
imshow(data, [], 'Colormap', cmap);
colorbar;
caption = sprintf('6000 gray levels with %d - %d in pseudocolor', row1, row2);
title(caption, 'FontSize', 15);

[John Doe]

Thank you. This is a much neater way of doing it.

I can add annotations using the title, but how would I go about using

insertText

to add text in an artitary position?

[Walter Roberson]

rgb = ind2rgb(data, cmap);
title_x = 100; title_y = 350;
rgbtitled = insertText(rgb, [title_x, title_y], 'The rain in Spain');

[Image Analyst]

insertText() burns the text into the image, changing the pixel values. text() places the text into the overlay, not altering the pixel values. Use whichever you want.

Answer 2

[Image Analyst]

Yes, you can use ind2rgb() to make the rgb image. Then you don't have to display it if you don't want to, or you could display it with imagesc() or image() instead of imshow().

[John Doe]

I already tried using ind2rgb(), but I am unable to set the range like I can in imshow. does ind2rgb include a function similiar to 'DisplayRange'?

[Walter Roberson]

See rescale() or the older mat2gray()

[John Doe]

Thanks for the input, unfortunately thats not what i'm looking for unless i'm missing something.

I want the colormap to only look at interval of pixels with an intensity of 0 to 6000, not change the pixel values themselves.

Instead of seeing a heatmap, my entire image turns white, although with the proper annotations

rescale_image = rescale(image,0,6000);
image = ind2rgb(rescale_image,CustomColormap);
text = 'custom text'
Annotated = insertText(image,[20 20; 20 20],text,'FontSize',12);
Annotated_title = insertText(Annotated,[20 1100],gif_filename,'FontSize',10);
imshow(Annotated_title)
colorbar

[Image Analyst]

"want to have a colormap on the image for further analysis on the picture such as adding annotations etc.

The only way I found that works, is by using imshow. The problem is that I don't want to display the image."

Then don't call imshow(). A colormap only applies if the image is shown. If the image is not shown, the image remains as a grayscale (indexed) image. If you want a color image instead, then you need to use ind2rgb(). You can do that without displaying it. If you want to burn annotations into the image, you can use insertText() or insertShape() if you have the Computer Vision Toolbox, which it looks like you do.

If you do call imshow((), call it with [] because your image is using a weird range:

imshow(Annotated_title, []);

Don't use image as the name of your image because that is the name of a built in MATLAB function and that could lead to errors.

rescale() like you used it maps the min of your image to 0 and the max of your image to 6000. You said you are "interested in a narrow band betwen 0 and 6000 pixel intensity" which means that your min and max are well outside that range. If you want that range exactly to scale you need to use the inputMax option to set everything above 6000 to 6000, otherwise something at 6000 may well end up at 500 or something way smaller than 6000

rescale_image = rescale(image,0,6000, 'InputMax', 6000);

Finally, I'm not sure how well imshow() handles floating point RGB images outside the range of 0-1 so you might want to cast the result to uint16 before calling imshow(), if you do end up calling imshow() which I know you said you don't want to do.

[Walter Roberson]

Floating point images are expected to be in the range 0 to 1, and not just by imshow()

[Walter Roberson]

... Is the idea that you want a colormap to be applied selectively to values in the range 0 to 6000, and that all other values are to be left as-is ?

If so then there are a couple of ways I can think of to do that:

1. colormap the entire range. Some values will map to themselves. Colormaps can have up to 65535 entries (used to be 256 entries for Windows.)
2. For display purposes only, display the image once with original colormap in an axes. Create a second axes and display the image again in exactly the same place, with caxis([0 6000]) and AlphaData set to 0 for all values outside the range 0 to 6000 (and 1.0 for values inside the range) so that the parts outside the range are transparent and show through to the original image. You need two axes for this because colormaps are per-axes, not per-image.
3. Since you say you do not want to display the image, run three interp1 calls, once for each color channel. The low value would be the lowest value possible for the data; the second lowest would be 0, then all the values where you "change direction" in linear color mapping. The second inputs would be the color information to map to.
4. Look up table, such as https://www.mathworks.com/matlabcentral/fileexchange/43004-imlut-img-lut-kind-order-colorscheme

[John Doe]

"... Is the idea that you want a colormap to be applied selectively to values in the range 0 to 6000, and that all other values are to be left as-is ?"

I see there's some confusion about exactly what I'm trying to achieve. but essentially that is what I want to achieve, except there's no other values.

my picture grayscale 16 bit picture only has pixel intensites between 0 and 6000. No need to rescale.

I want to visualize the difference between these using a colormap that can show difference between, for example, 3000 and 3005, and afterwards add text to the colormap using insertText.

I found an alterative to imshow but it does not help much since imagesc also displays the image and i'm unable to use insertText on this

colormap(PlusMinus);
chart = imagesc(BlurJ);
ax = gca;
ax.CLim = [0 6000];
axis off

Now I want to use insertText but to be honest have no idea what exactly to annotate. it gives me an error when i try gca (Expected input number 1, I, to be one of these types:uint8, uint16, int16, double, single Instead its type was matlab.graphics.axis.Axes.), and adding the annotation before the colormap seem to change values.

colormap(PlusMinus);
chart = imagesc(BlurJ);
ax = gca;
ax.CLim = [0 6000];
insertText(gca,[20 1100],'hallo','FontSize',10);
axis off

when I try using ind2rgb() it does not apply colormap correctly. it seems to assume most of the intensity is out of range.

BlurJ = ind2rgb(BlurJ,PlusMinus);
chart = image(BlurJ);
ax = gca;
ax.CLim = [0 6000];
axis off

[Walter Roberson]

I want to visualize the difference between these using a colormap that can show difference between, for example, 3000 and 3005, and afterwards add text to the colormap using insertText.

Generally speaking, you can use ind2rgb() or other forms of lookup table in order to create a color array representing the data. Then you can use Computer Vision insertText() to draw text into that array if desired.

The question at this time would be how you want to represent the color scheme.

See for example https://www.mathworks.com/matlabcentral/fileexchange/69470-custom-colormap which permits you to define relative positions (on the range 0 to 1) and corresponding colors, and will interpolate between them to the number of locations you specify. If you specified 6001 then you could use ind2rgb() with the map you had created. Otherwise you could use three interp1 calls.

Indeed, if you have a list of positions and rgb components and are willing to interpolate between them, then you can use three interp1() calls directly without using the FEX contribution. The first parameter would be the positions of change; the second parameter would be the color component value at those locations; the third parameter would be the image value.