Converting grayscale image to indexed

I have a grayscale image with 3 random values signified by integers 91, 123, 255. When I use gray2ind(I, 3) I get only 2 indices: 1 and 2. But when I use gray2ind(I, 2) I get 0 and 1 as expected. How come? I would expect gray2ind(I, 3) to give me 0, 1 and 2.

I do have a colormap array which I am planning to later use, which has 3 individual colors, such that after I would properly convert this to 0, 1 and 2 indices I'd just do imshow(gray2ind(I, 3), colors).

Thanks

Answer 1

[Image Analyst]

Don't worry about it. You can use a grayscale image as an indexed image as-is. Just set up the colormap based on the grayscale and set the map for whatever colors you want for the special values.

grayImage = imread('pout.tif');
% Initialize a normal colormap for a normal gray scale image.
cmap = gray(256);
% Set up special colors for 91, 123, 255.
cmap(92, :) = [1, 1, 0]; % Gray scale 91 will show up as yellow.
cmap(124, :) = [1, 0, 1]; % Gray scale 123 will show up as magenta.
cmap(256, :) = [1, 0, 0]; % Gray scale 255 will show up as red.
imshow(grayImage, cmap);
colorbar
% Enlarge figure to full screen.
set(gcf, 'Units', 'Normalized', 'OuterPosition', [0, 0.04, 1, 0.96]);

However, note that the index (row) of the colormap you want to use is one more than the gray level, because row 1 controls the color of gray level 0, row 2 controls the color of gray level 1, and so on until row 256 controls the color of gray level 255.

[Micha]

Thank you for clarifying this use. I wasn't aware that the values in the pixels correspond to their position in the colormap, from whence the colors are taken.

As a follow-up, I was wondering if it is possible to accomplish this without having to set cmap(92, :) = [1, 1, 0]; cmap(124, :) = [1, 0, 1]; cmap(256, :) = [1, 0, 0]; as you have shown, but rather have it in order - so that the order of the grayscale values would correspond to the order in the cmap, so that I can initially have cmap = [1 1 0; 1 0 1; 1 0 0]; this would have worked if I converted 91, 123, 255 in the original image to 0, 1, 2, which is the reason for my initial question.

[Image Analyst]

You can but it's a lot of extra code that only complicates things. Why would you want to do that instead of the very simple method I gave you?

[Micha]

Because I have other images that are indexed to begin with, so this method is compatible with them. The instance I described here is the case with some images, so it's more convenient to convert these to 0 1 2

[Image Analyst]

Try this:

grayImage = imread('pout.tif');
% Quantize into 3 gray levels.
qImage = imquantize(grayImage, [91, 123, 255]);
unique(qImage)
% Set up a color map for color 1 valid for GL <=91,
% color 2 valid for GL >= 92 up to GL <= 123,
% and color 3 valid for GL >= 124 up to GL <= 255.
cmap = [1, 1, 0;
  1, 0, 1;
  1, 0, 0];
imshow(qImage, cmap);
colorbar
% Enlarge figure to full screen.
set(gcf, 'Units', 'Normalized', 'OuterPosition', [0, 0.04, 1, 0.96]);

Answer 2

[Micha]

Thanks, imquantize was what I was looking for.