Part 2: 2D Median Filter With Different Window Size

MATLAB CODE:

clear all

%READ AN IMAGE

I = imread('zebra.jpg');

display(size(I));

%CONVERT RGB IMAGE INTO GRAYSCALE

A = rgb2gray(I);

%ADD SALT AND PEPPER NOISE TO THE GRAYSCALE IMAGE

A = imnoise(A,'Salt & pepper',0.1);

figure,imshow(A);title('IMAGE WITH SALT AND PEPPER NOISE');

%DEFINE THE WINDOW SIZE MXN

M=5;

N=5;

%PAD THE MATRIX WITH ZEROS ON ALL SIDES

modifyA=padarray(A,[floor(M/2),floor(N/2)]);

figure,imshow(uint8(modifyA)); title('PADDED WITH ZEROS');

B = zeros([size(A,1) size(A,2)]);

med_indx = round((M*N)/2); %MEDIAN INDEX

for i=1:size(modifyA,1)-(M-1)

    for j=1:size(modifyA,2)-(N-1)

       

       

        temp=modifyA(i:i+(M-1),j:j+(N-1),:);

        tmp_sort = sort(temp(:));%tmp(:) converts 2D matrix to 1D matrix

        B(i,j) = tmp_sort(med_indx);

      

      

       

    end

end

 %CONVERT THE IMAGE TO UINT8 FORMAT.

B=uint8(B);

figure,imshow(B);

title('IMAGE AFTER MEDIAN FILTERING');

EXPLANATION:

·        The image is read into the matrix A.  

·        The image is color matrix that contains Red, Green and Blue channels so ‘rgb2gray()’ command is used to convert RGB image to grayscale.

      Learn how to convert RGB image to grayscale image

·        Salt and pepper noise is added to the image.  Learn how to add 'salt and pepper noise to an image'.

·        Define the window size MxN example: 3x3, 7x7, 5x3

·        Pad the image with zeros on all sides. This is done to perform the filtering on the border pixels. Learn howto pad with zeros using MATLAB built_in function padarray.

·        Median is the middle point of the series. The index that is obtained by dividing the total number of elements in a window by 2 gives the position.

·        A sliding window of size M x N is used and the elements in the window are sorted and the middle element from the sorted array is chosen.

EXPLANATION:

·        The image filtered with 3x3 window still contains salt and pepper noise but the edges are still sharp.

·        The image filtered with 5x5 window  contains less noise compared to 3x3 window filter and edges are smoothed a bit

·        The image filtered with 7x7 window is free of noise but the edges are smoothed to certain extent.