Code issue with custom imhist() function

Hi everyone, recently my classes required me to write a function to create a histogram of a greyscale image without using the built in function imhist().

I came up with this but there's something I don't understand.

function out_hist = myhistogram(image)
    [rows, cols] = size(image);
    out_hist = zeros(256, 1);
    
    for i = 1 : rows
        for j = 1 : cols
            value = image(i, j);
            % code here
        end
    end
end

With just this expression (to replace in "code here") when you encounter a 255 value it gets added to the 255th position instead of the 256th.

out_hist(value + 1) = out_hist(value + 1) + 1;

While this, which does the exact same thing, works perfectly.

if (value == 255)
    out_hist(256) = out_hist(256) + 1;
else
    out_hist(value + 1) = out_hist(value + 1) + 1;
end

Now I want to ask: am I missing something? Don't they produce the excact same result?

Thanks in advance for the answers.

Answer 1

[Jan]

You observe the effect of a saturated integer class:

x = uint8(0);
x = x + 254

x = uint8 254

x = x + 1

x = uint8 255

x = x + 1  % !!!

x = uint8 255

An UINT8 cannot contain a value greater than 255. In Matlab all greater values are stored as maximum value. The same effect occurs for too small values:

x = x - 256

x = uint8 0

If the image is stored as UINT8, value=image(i,j) creates value as an UINT8 also.

A solution is to use a class for the index, which has a higher capability:

value = double(image(i, j));
out_hist(value + 1) = out_hist(value + 1) + 1;

The manual treatment of this exception is fine also:

if (value == 255)
    out_hist(256) = out_hist(256) + 1;    

Here the number 256 is treated as double automatically, because this is the defult in Matlab.

[Davide]

Oh I see. Thanks for the explanation.

Just to clarify: when I do value+1 as the out_hist argument, does a temp variable get created to store the addition ? Does it automatically gets set to an uint8? Is this the reason it was capped at 255?

[Steven Lord]

Your image variable (which you may want to rename, BTW, as image already has a meaning in MATLAB) is an array of class uint8. Adding a uint8 (the value element you extracted from image) and a scalar double (1) results in a uint8. See the section on arithmetic operators on this documentation page.

[Jan]

@Davide: Fell free to try it:

x = uint8(1);
class(x)
class(1)
class(x + 1)
class(1 + x)
class(1 + 17)
% etc.

Or maybe less useful: yes, yes, yes.

[Davide]

Ok, thanks again. Will definitely try it and do some experiments.