Why does the "graythresh" function return 0 on input images of type "double"?

The documentation for the "graythresh" function says that it accepts images of type "uint8", and of type "double".

For "uint8", the "graythresh" function works as expected, as follows : 

>> I = imread('coins.png');
>> level = graythresh(I)

level = 
0.4941

>>imshow(im2bw(I, level));

However, when the input image is of type "double", the level is 0 as follows :

>> I = imread('coins.png');
>> level = graythresh(double(I))

level = 
0

Why does this happen even though the "graythresh" function can accept inputs of type "double"?

Answer 1

[MathWorks Support Team]

The "graythresh" function assumes that if the input image is of type "double", then the image is scaled between 0 and 1. 

So, in order to correctly use the "graythresh" function with "double" images, scale the input image between 0 and 1 as follows : 

 

I = imread('coins.png');
dbI = double(I);
dbI = dbI/max(max(dbI));
level = graythresh(dbI);
imshow(im2bw(I, level));

[Anand]

The alternate would be to use im2double instead of just double. This function does the necessary scaling.

Also, imbinarize is the preferred function to use instead of im2bw.

I = imread('coins.png');
dbI = im2double(I);
level = graythresh(dbI);
imshow(imbinarize(I, level))

In fact, you don't even need to separately call graythresh. The imbinarize function does this internally.

I = imread('coins.png');
dbI = im2double(I);
imshow(imbinarize(I));