why doesn't Graythresh work with the format "double"?

I have an image (gray scale, 0-255) in the form of "uint8". I only need a part of it and as such I do the following:

Part_of_the_Image=original_Image(1:H,1:W);

the above, gives me an Image in the form "double". If I plot it (via imagesc and setting the colormap to gray), I would get a "white" image in the figure. In addition, the "graythresh" function does not work on it either (would give 0 as the threshold).

All of the problems are NOT encountered if I convert "Part_of_the_Image" to uint8.

I am just wondering why "double" yields to such problems!

Any comment would be appreciated.

Answer 1

[Image Analyst]

If your original_Image is truly uint8, then Part_of_the_Image will also be uint8. It will not convert it. Use class() or whos() to check. Post proof if you think otherwise. Here's my proof:

original_Image = 255 * ones(200, 'uint8');
H = 15;
W = 9;
Part_of_the_Image = original_Image(1:H,1:W);

whos original_Image
whos Part_of_the_Image

In the command window:

    Name                  Size             Bytes  Class    Attributes

    original_Image      200x200            40000  uint8              

    Name                    Size            Bytes  Class    Attributes

    Part_of_the_Image      15x9               135  uint8

If an image is floating point, it's expected to be in the range 0-1 and pixels outside that range will be black or white. If you have some arbitrary range that's different than that, you can display the full dynamic range with [] option in imshow():

imshow(doubleImage, []);

[ramin bba]

The problem was I predefined the variable as zeros(), which makes it "double".

however, I still do not know why some functions (Graythresh or imagesc) do not work when the format is double! in the documentation it is stated that the input can be of the form "double".

[Image Analyst]

You can pass in 'uint8' into zeros() and ones() like I did to make it be uint8.

graythresh() can take a double but, like with all image processing toolbox functions, it needs to be in the range 0-1. It's a pain so I never use it.

I don't see anything in the help about imagesc() taking doubles. If this answers your questions, please vote for and Accept the Answer.