How to write own standard deviation function.

I have read the color image. Then separated RGB values into three different arrays. After that I have written my own function to calculate standard deviation function for each color component. But when I execute my own written function and in built function I got different values? What is wrong int it?

without inbuilt function

im = imread('D:\im112.jpg');
R=im(:,:,1)
[r,c]=size(R);
totmean=sum(R(:))/(r*c);
totdiff=(R-totmean).^2;
totsum=sum(totdiff(:));
nele=(r*c)-1;
totvar=totsum/nele;
totstd=sqrt(totvar);
display(totstd);

Using inbuilt functio

stdr=std(double(R(:)))

[Adam]

Is your image in integer format? You convert to double to call the builtin, but seemingly not for your own code. I don't know off the top of my head what the result of all those operations is on integers.

Answer 1

[Guillaume]

Adam hit the nail on the head in his comment. You're reading a jpg image so most likely the im array and thus R is of type uint8 (8 bit integer). The range for uint8 value is 0 to 255. If any operation involving uint8 overflows the result is clamped to 255 and similarly underflows are clamped to 0.

Therefore, it's highly likely that the result of sum(R(:)) is clamped to 255. Then you do a division. Because the result is still going to be uint8 the result is going to be rounded down to the nearest integer (most likely 0, since 255 divided by the image area is probably less than 1).

To solve this, you indeed need to do the same as you've done for std. First convert your R to double, then all operations will be performed correctly.

[anu]

Thanks. When I converted R to double, it gives me correct answer.

Answer 2

[Andrei Bobrov]

n = numel(R);
yourstd = sqrt(sum((R(:) - sum(R(:))/n).^2)/(n - 1));

Answer 3

[Image Analyst]

Why not simply use std2() - the built in function meant for this????

img = imread('moon.tif');
s = std2(img) % No casting to double needed.