Closet-neighbor-image-filtering

%clearing command window, workspace and image windows
clc;
clear all;
close all;

%loading image
original_image = imread('grayscale_cat.jpg');

%adding noise, using gaussian noise equation
%of mean 10 and variance 20
mean = 10;
sigma = sqrt(20);
original_image= double(original_image);
an = mean + (randn(size(original_image))*sigma);
noisy_image = original_image + an;


[m,n]=size(noisy_image);

%usng a mask of 3-by-3

mask_matrix = [1 2 1; 2 4 2; 1 2 1];

%z=ones(3);

[p,q] = size(mask_matrix);

w=1:p;
x = round(median(w));
%{
below, we are making noisy matrix's corner's element all zero so that we
can use 3-by-3 kernel with ease for corner elements
%}

noisy_image_with_extra_zeros = zeros(m+2*(x-1),n+2*(x-1));

for i=x:(m+(x-1))

     for j=x:(n+(x-1))

        noisy_image_with_extra_zeros(i,j) = uint8(noisy_image(i-(x-1),j-(x-1)));

     end

end

figure, imshow(uint8(noisy_image_with_extra_zeros))
title('noisy-image with extra zeros around');

x=0;

y=0;
%these loops map the noisy_image with zeros to the kernel so that summation
%can occur after multiplication between them
for i=1:m

    for j=1:n

        for k=1:p

            for l=1:q

                summation(k,l)= noisy_image_with_extra_zeros(i+x,j+y)*mask_matrix(k,l);
                y=y+1;
            end

            y=0;

            x=x+1;
        end

        x=0;
       %here we are finding the closest pixel to consider for our outcome
       %image
        to_be_sorted= abs(summation - double(noisy_image(i,j)));

        sorted_vector = sort(to_be_sorted(:));

        filtered_image(i,j)=(sum(sorted_vector(1:5)));
        summation=[];

    end
end

figure, imshow(uint8(filtered_image))
title('Filtered Image by taking 5 closest point');