sliding window

How to apply a 3x3 sliding window for an image while the pixel in the center of the window should have the mean value of the window.

Answer 1

[Ben Mitch]

I think you're after filter2(). Try

mn = filter2(ones(3), im);

[Gova ReDDy]

sorry..i didnt get you can xplain it clearly

[Jan]

Read "doc filter2". I assume you need a further normalization: "filter2(ones(3)/9, im)".

[Ben Mitch]

assuming your image comes from a file (if not, the process is much the same)...

im = imread('my_image.png');

im = double(im);

im = mean(im, 3); % convert to greyscale, if is RGB

b = ones(3)/9;

im = filter2(b, im);

imshow(im);

if you need to work in RGB rather than greyscale, you could use something like...

for d = 1:3

im(:, :, d) = filter2(b, im(:, :, d));

end

[Gova ReDDy]

danx..bt i want to store the mean value of the 3x3 sliding window in the centre pixel and the window should slide to the next 3x3 position

[Image Analyst]

But that's exactly what Ben, Jan, and Sean de's code does! Did you look up the help for filter2()? It is basically conv2() (which is highly optimized) with the 'same' option. (I always use conv2 rather than filter2.) Write a test with a small image to convince yourself.

Answer 2

[Jan]

What is "the next 3x3 position"? Will the resulting picture have a width and size reduced by 3? If so, this can achieve the calculation efficiently: FEX: BlockMean

[Gova ReDDy]

Not the same actually image size is 256*256 pixels and the sliding window size is 3*3 each centre pixel value of window should have the window mean value,"the next 3*3 window " means the window should slide along all the pixels of the image.

[Image Analyst]

Patan, how many times do we have to say it. CONV2 AND IMFILTER WILL DO WHAT YOU'RE ASKING!!!!

Answer 3

[Sean de Wolski]

Imean_window = conv2(double(I),ones(3)/9,'valid');