How can I extract the white pixels of an image?

I want to get some help to extract the white part of an image.

For example, in this image, I want to extract the area shown in the red retangular.

I want the area to be substracted is at least 20 *20pixel width and length, like shown in the example above. It means that any small white areas, such as a single white point, is ignored.

Can I get some help?

Thanks a lot!

Answer 1

[Image Analyst]

Simply threshold to create a binary image of where "white" is. Then call beareaopen() to remove blobs smaller than a certain size. Then call regionprops() to get the bounding box of the blob you want to extract. Then call imcrop() to cut it out and display it separately. See attached code, below this image it creates:

[Chen Zhu]

Thanks for your help! Your code works well!

I have a question. Why you choose 200 as the threshold for each channel? Is 200 representing white?

Thanks!

[Image Analyst]

Pick whatever you want. That value seemed to give a white spot about the size of what you wanted. Higher thresholds give smaller spots.

Answer 2

[John Chilleri]

Hello,

Assuming your image is of size m x n x 3, where the stored values range up to 255, you could use the following simple approach:

% Assume img is your m x n x 3 image

img = double(img);

[m, n, ~] = size(img);

% Step one, constrain to lower left, as we don't want to lose her shoes or the upper left:

m1 = round(2*m/3);
n1 = round(1*n/3);
submat = img(m1:m,1:n1,:); % fit this accordingly, I just guestimated

% Determine location of white pixels using threshold and change:

for i = 1:size(submat,1)
    for j = 1:size(submat,2)
        if (sum(submat(i,j,:)) > 3*240)
             % White pixel - do what you want to original image
             img(m1-1+i,j,:) = [0 0 0]; % make it black, for example
        end
    end
end

img = uint8(img);

Using the above code on a screen shot of your image with a threshold of 240, I get this:

Note: the image is cut off because I snapped a quick screen shot that wasn't carefully cropped.

There are probably much better methods, but this is a simple approach.

Hope this helps!

[Chen Zhu]

Thanks for your help! I want my code to work for any pictures, so there is no need to set the constrain at step one.

I tried to change the code to

for i = 1:size(img,1)
    for j = 1:size(img,2)
        if (norm(img(i,j,:)) > 240)
             % White pixel - do what you want to original image
             img(i,j,:) = 2;
        end
    end
end

However, I got the error: Undefined function 'norm' for input arguments of type 'uint8'.

In addition, I want to extract the white areas from the image. I want the area to be substracted is at least 20 *20pixel width and length, like shown in the example above.

Thanks!

[John Chilleri]

Finally finished editing my code, sorry for the delay, tell me how things work now!

I find that a threshold of 3*210 seems to get a majority of the reflection (if not too much):

[John Chilleri]

In order to extract an area, you could make the submat = img, and set equal to black if less than threshold, else set to white. Then you have a white image with black splotches that you can then select (if they're big enough) to extract.

Answer 3

[chandu priya]

img = double(img);

[m, n, ~] = size(img);

% Step one, constrain to lower left, as we don't want to lose her shoes or the upper left:

m1 = round(2*m/3);

n1 = round(1*n/3);

submat = img(m1:m,1:n1,:); % fit this accordingly, I just guestimated

% Determine location of white pixels using threshold and change:

for i = 1:size(submat,1)

for j = 1:size(submat,2)

if (sum(submat(i,j,:)) > 3*240)

% White pixel - do what you want to original image

img(m1-1+i,j,:) = [0 0 0]; % make it black, for example

end

end

end

img = uint8(img);