(MIP Lab) Noise Removal in Digital Image Processing
(MIP Lab) Noise Removal in Digital Image Processing
Image Processing
Image Restoration and Reconstruction
What is Image Restoration?
It is the purpose of recovering an image that has been degraded by applying restoration techniques in modelling the
degradation and applying inverse process to recover the image.
What is the need for Image Restoration?
Whereby, f(x,y) is the image, H(x,y) is the degradation function, (x,y) is the noise added to the image, g(x,y) is the image
መ
that is added with noise and 𝑓(x,y) is the restored image.
• Gaussian Noise
Where, where z represents intensity, z is the mean (average) value of 𝑧,ҧ and is its standard deviation. Variance will be the
square of standard deviation, i.e., 2. When the values of z are in the range 𝑧ҧ ± , the probability is about 0.68 (almost 70%)
and when the values are in the range 𝑧ҧ ± 2, the probability is about 0.95 (95%).
• Rayleigh Noise
The mean and variance of z when this random variable is characterized by a Rayleigh PDF are:
where the parameters are such that a > b, b is a positive integer, and “!” indicates factorial. The mean and variance of z are:
Mean – Variance –
• Exponential Noise
Mean – Variance –
Mean – Variance –
k represents the number of bits used to represent the intensity values in a digital image, with the range being [0,2k - 1] or
[0-255] in the case of an 8-bit image.
• Order Statistics
Median filter
Max-Min filter
Midpoint filter
Alpha-trimmed filter
• Adaptive filters
• Inverse Filter
• Minimum Mean Square Error (Wiener filter)
• Median Filter
The best-known order-statistic filter in image processing is the median filter, which, as its name implies, replaces the value
of a pixel by the median of the intensity levels in a predefined neighborhood of that pixel:
Sxy is a subimage (neighborhood) centered on point (x, y). It is best used for filtering speckle and salt & pepper noises.
• Wiener Filter
Where E{} is the expected value of the argument and e2 is the error measurement.
The minimum mean square error in the frequency domain is given by:
Code:
I = imread('coins.png');
figure; imshow(I); title ('Original Image');
figure; subplot(3,2,1);
gauss_noise = imnoise(I,'gaussian',0);
imshow(gauss_noise); title('Gaussian Noise');
subplot(3,2,2)
wien = wiener2(gauss_noise,[5 5]);
imshow(wien); title('Wiener filtering of Gaussian noise');
subplot(3,2,3);
saltpep = imnoise(I,'salt & pepper',0.05);
imshow(saltpep); title('Salt & Pepper Noise');
subplot(3,2,4);
saltpep_filt = medfilt2(saltpep);
imshow(saltpep_filt); title('Median Filtering of salt & pepper noise');
subplot(3,2,5);
speckle = imnoise(I,'speckle',0.05);
imshow(speckle); title('Speckle Noise');
subplot(3,2,6);
speckle_filt = medfilt2(speckle);
imshow(speckle_filt); title('Median Filtering of speckle noise');
RESULTS