How to convert values of 0 to NaN

When I plot my graphs using imagesc, how do I conver the values of 0 to NaN?

Currently, the colorbar displays 10^0 as the lowest, but if the value is 0, it still displays this blue color. I want the graph to display white if the value is equal to 0.

Answer 1

[Matt J]

NaNs will not appear bright on an image display. You would have to replace them with a large value.

Image(Image==0)=brightValue;

Answer 2

[Walter Roberson]

Leave them the way they are now, but set the axes color to white. Tell imagesc() tol make NaN values transparent, so those locations would show the white background underneath.

im = imread('cameraman.tif');
figure
imagesc(im);
colormap(gray);
title('original');

nim = im2double(im);
nim(randi(numel(nim), 1, 3000)) = nan;
figure
imagesc(nim, 'alphadata', 0+~isnan(nim));
colormap(gray)
set(gca, 'color', 'r');
title('nan shows color underneath')

[Eric]

I don't quite understand what this does:

nim = im2double(im);

nim(randi(numel(nim), 1, 3000)) = nan;

Do you mind explaining to me?

[Walter Roberson]

The image that was read for example purposes was uint8. uint8 cannot represent NaN, so I converted the image to double precision. Then I added in some nan values.

I needed an image that contained nan values in order to show how to use imagesc and axes properties to allow you to see through an image.

[Eric]

Is 0+~isnan(nim) what converts the 0 squares to NaN?

[Walter Roberson]

@Eric

Your original question involved having data that has NaN inside it.

isnan(im) would respond with a logical array that is true (==1) for all NaN values and false (== 0) for non-nan values.

~isnan(im) would then be a logical array that is false (== 0) for all NaN values and true (== 1) for non-nan values.

0+~isnan(im) leaves the numeric values the same but converts to double precision. So you would have 0 (double) for all NaN values and 1 (double) for all non-NaN values.

Those 0 and 1 (double) are used as AlphaData values. Alpha values of 1 (in this case, meaning was not NaN) tell image() objects to be fully opaque -- so the color implied by the (non-NaN) data will be used at that location. And alpha values of 0 (in this case meaning the data was NaN) tell image() objects to make the pixel fully transparent, allowing whatever is underneath to be seen through at that location.