Question

2) This project aims to reduce image noise using mask operation. a) Use imnoise MATLAB function to add different noises to the input image. b) Investigate the usefulness of each of the mean, median and Gaussian filtering for removing different types and levels of image noise (Hint: use fspecial and imfilter commands in MATLAB). c) Investigate the effect of different 3x3 averaging masks. d) Investigate the effect of filter size on the computation time. Consider Gaussian filter and change its size and record the corresponding operation time. Does the timing increase linearly or not? Why is this? (Hint: to measure the time use tic and toc MATLAB comments).

Answer 1

ANSWER:

• I have provided the properly commented code in both text and image format so you can easily copy the code as well as check for correct indentation.
• I have provided the output image of the code so you can easily cross-check for the correct output of the code.
• Have a nice and healthy day!!

CODE TEXT

% Reading sample.JPG image

img = imread("sample.JPG");

% Grayscaling the image

input_img = rgb2gray(img);

%% a. Adding different types of noises using imnoise

figure(1);

% Showing original image

subplot(2,2,1);

imshow(input_img);

title('Original Image');

% Adding gaussian noise to image

gauss_noise = imnoise(input_img,'gaussian');

% Showing gauss_noise image

subplot(2,2,2);

imshow(gauss_noise);

title('gaussian noise image');

% Adding poisson noise to image

poisson_noise = imnoise(input_img,'poisson');

% Showing poisson noise image

subplot(2,2,3);

imshow(poisson_noise);

title('poisson noise image');

% Adding salt & pepper noise to image

snp_noise = imnoise(input_img,'salt & pepper');

% Showing salt & pepper noise image

subplot(2,2,4);

imshow(snp_noise);

title('salt & pepper noise image');

%% b. Investigating usefullness of different filters

figure(2);

% Creating a cell of different noised images

noise_images = {gauss_noise, poisson_noise, snp_noise};

noise_images_labels=["gaussian", "poisson", "salt & pepper"];

% defining empty vector to store snr values for different noise and

% filterings

filter_snrs=[];

% looping for each type of noise image and applying different filters

for i=1:length(noise_images)

% defining empty vector to store snr values for different filtering

filter_snr=[];

% fetching single noised image from cell

noise_image = noise_images{i};

% ploting noise_image

subplot(3,4,(i-1)*4+1);

imshow(noise_image);

title(sprintf('%s noise Image',noise_images_labels(i)));

% a. applying mean/average filter

mean_filter = fspecial('average');

mean_image = imfilter(noise_image, mean_filter);

% finding snr

snr_val = snr(double(input_img),double(mean_image));

filter_snr = [filter_snr snr_val];

% ploting

subplot(3,4,(i-1)*4+2);

imshow(mean_image);

title('Mean filter');

% b. applying median filter, using medfilt2

median_image = medfilt2(noise_image);

% finding snr

snr_val = snr(double(input_img),double(median_image));

filter_snr = [filter_snr snr_val];

% ploting

subplot(3,4,(i-1)*4+3);

imshow(median_image);

title('Median filter');

% c. applying gaussian filter

gaussian_filter = fspecial('gaussian');

gaussian_image = imfilter(noise_image, gaussian_filter);

% finding snr

snr_val = snr(double(input_img),double(gaussian_image));

filter_snr = [filter_snr snr_val];

% ploting

subplot(3,4,(i-1)*4+4);

imshow(gaussian_image);

title('Gaussian filter');

% appending the snr_values to complete list

filter_snrs=[filter_snrs; filter_snr];

end

%% b&c. ploting snr to investigate the effect of different filters on

% different types of noise

figure(3);

for i=1:size(filter_snrs,1)

subplot(3,1,i);

plot([1],filter_snrs(i,1),'*');

hold on;

plot([2],filter_snrs(i,2),'*');

hold on;

plot([3],filter_snrs(i,3),'*');

legend('Mean filter','Median filter','Gaussian filter');

title(sprintf('SNR Values for Different filters on %s noise Image', ...

noise_images_labels(i)));

ylabel('SNR');

end

%% d. Investigating effect of kernel size on computational time

kernel_sizes=3:2:30;

% defining vector to store time

time_taken=[]

for k=kernel_sizes

% fetching map of size k

gaussian_filter = fspecial('gaussian',k);

% initializing tic

tic;

% applying kernel

gaussian_image = imfilter(noise_image, gaussian_filter);

% checking time

time_taken = [time_taken, toc];

end

% ploting graph Kernel size vs time

figure(4);

plot(kernel_sizes,time_taken);

xlabel('Kernel Size');

ylabel('Computational Time');

title('Kernel Size vs Computational Time');

CODE IMAGE

OUTPUT IMAGE

a.

b .

c.

d.