Calculate the mean value of border pixels in image and minus that from the whole image

Hello,

I am looking for what I am hoping is quite a simple solution.

I have an image that at the moment is 200 by 200 but has the ability to change size (lengh and width but at an equal rate).

What I need: To look at only the outside 5 pixels or so of the whole image ( so a 5 pixel wide band around the image) and calculate the mean pixel value.

I then need to subtract that value from each pixel in the image.

Any help is greatly appreciated,

Ollie.

Answer 1

[Jan]

img = rand(30, 60, 3);  % Assuming RGB images;
border = 5;

siz  = size(img);
mask = false(siz(1), siz(2));
mask(1:border, :) = true;
mask(:, 1:border) = true;
mask(siz(1)-border+1:siz(1), :) = true;
mask(:, siz(2)-border+1:siz(2)) = true;

tmp = reshape(img, [], 3);
meanBorder = mean(tmp(mask, :), 1);
img = img - reshape(meanBorder, 1, 1, 3);

% Cross-check: do the border pixels have a zero mean value now:
tmp = reshape(img, [], 3);
meanBorder = mean(tmp(mask, :), 1)

meanBorder = 1×3

1.0e+-15 * 0.0130 0.1765 -0.0405

Fine. But now some pixels have negative values.

[Oliver Horrobin]

Hi Jan, *edit as I mistyped the code originaly

Thankyou very much^^.

It is coming up with these errors when I input it into the code and run it with the image:

Error using reshape

Product of known dimensions, 3, not divisible into total number of elements,

40000

(In this part)

tmp = reshape(outff, [], 3);

I changed the size of the image so it was divisable by 3 and got this error:

The logical indices in position 1 contain a true value outside of the array

bounds.

(in this bit)

meanBorder = mean(tmp(mask, :), 1);

I'm not sure why that happens?

Also you think you could explain how this works a little please?

Thanks again,

Ollie

[Jan]

Maybe your input is not an RGB image, which are [M x N x 3] arrays.

If it is a grayscale image, which are [M x N] matrices, use:

siz  = size(img);
mask = false(siz(1), siz(2));
mask(1:border, :) = true;
mask(:, 1:border) = true;
mask(siz(1)-border+1:siz(1), :) = true;
mask(:, siz(2)-border+1:siz(2)) = true;

meanBorder = mean(img(mask));
img = img - meanBorder;

[Oliver Horrobin]

Hi Jan,

It must be a grayscale image. It works perfectly now.

Thankyou so much,

Ollie