How to apply average filter on image ?

I m working on image to apply average filter on it. It gives perfect result on array of matrix but not working on real image here is my code. what I m doing wrong ?

S = imread('15.jpg');

I = rgb2gray(S);

% I = [100 255 94 30 150; 53 176 255 198 140; 200 113 118 255 250; 100 255 150 160 12; 255 255 125 152 53];

I2 = padarray(I,[1 1],'replicate','both');

[x,y] = size(I2);

for i = 2:x-1

for j = 2:y-1

sum = 0;

for ii = i-1:i+1

for jj = j-1:j+1

sum = sum + I2(ii,jj);

end

end

I3(i,j) = floor(sum/9);

end

end

subplot(1,3,1);

imshow(I)

title('Original Image');

subplot(1,3,2);

imshow(I2)

title('1 Row & Column added to Image');

subplot(1,3,3);

imshow(I3)

title('Averge Filter Image');

Answer 1

[Image Analyst]

sum is a built-in function so don't use it as one of your variable names.

Use the built-in imfilter(), or conv2().

rgbImage = imread('peppers.png');
windowWidth = 21;
kernel = ones(windowWidth) / windowWidth .^ 2;
subplot(2, 1, 1);
imshow(rgbImage);
drawnow;

blurryImage = imfilter(rgbImage, kernel, 'replicate');
subplot(2, 1, 2);
imshow(blurryImage);

[khaikheat kheat]

windowWidth = 21;

kernel = ones(windowWidth) / windowWidth .^ 2;

May you explain these 2 lines? Im new in matalb. Thanks

[khaikheat kheat]

When I change the value in the windowWidth, 23,25 adn 45 respectively. It coming out different filtered image. Can you help me with it?

[Image Analyst]

Of course, as it should. And we can help if you explain your difficulties in a new question (not here in a comment to an Answer to Hanif's question).

[khaikheat kheat]

OK. I will create a new question

Answer 2

[David Fernandes]

Although @Image Analyst solution is correct, in fact, it does not answer to your question. And I commend you for trying to understand the algorithm and not only use some built in function that hides it from you (at least at first). :)

So, maybe a bit too late here goes my 50 cents. :)

The problem is that by default when you say: sum = 0, sum becomes a UINT8 which only accepts integers up to 255. Adding 9 UINT8 values will, almost for sure, overflow 255 and 255/9 is almost black.

So, you have to force a "bigger" type.

For instance:

for i = 2:x-1
    for j = 2:y-1
        sum = uint16(0); % change here
        for ii = i-1:i+1  
            for jj = j-1:j+1
                sum = sum + uint16(I(ii,jj)); % and here
            end
        end
        I2(i,j) = floor(sum/9);
    end
end

[Image Analyst]

Since sum is a built-in function, you should call it "theSum" so that you do not overwrite the built-in sum() function.

Also you seem to have non-standard nomenclature where you have i, the first/row index, be the x direction, and j, the second/column index, go up to "y-1". Remember, images are addressed as (row, column), not as (x,y) so you should say I2(j, i) and I(jj, ii) if you're using the usual convention of x being the horizontal/columns direction, and y being the vertical/rows direction.

But other than that, it should work for a gray scale image.

[David Fernandes]

Well, you're right, thank you for your comments. Although my only concern was to try to explain why the resulting image was not as expected and to make the minimum changes to the original code in order to fix it.