Help with salt_and_p​epper_nois​e_removal.​m

I found a matlab code to remove salt and pepper noise from a color image written by Image analyst in here . I have two question regarding to the code.

1. How to skip the first and last column and row of the image from median filter?

2.

% Find the noise in the green.

noiseImage = (greenChannel == 0 | greenChannel == 255);

% Get rid of the noise in the green by replacing with median.

noiseFreeGreen = greenChannel;

noiseFreeGreen(noiseImage) = greenMF(noiseImage);

What is the value of variable named noiseImage? Is it a matrix same size as the parent image but has value only where greenchannel is 0 and 255?

[R. B.K.]

For my 1st question's solution I tried to add restriction in median filter with this method.

% Median Filter the channels:
redMF = medfilt2(redChannel(2:rows-1, 2:columns-1), [3 3]);
greenMF = medfilt2(greenChannel(2:rows-1, 2:columns-1), [3 3]);
blueMF = medfilt2(blueChannel(2:rows-1, 2:columns-1), [3 3]);

But this gives me an error:

??? Index exceeds matrix dimensions.

Error in ==> salt_and_pepper_denoising at 88
noiseFreeRed(noiseImage) = redMF(noiseImage);

Answer 1

[Mehmet Killioglu]

1. You already proposed a solution but you're changing redMF's size. You should merge with redChannel's first and last columns and rows.

% Median Filter the channels:
redMF = medfilt2(redChannel(2:rows-1, 2:columns-1), [3 3]);
greenMF = medfilt2(greenChannel(2:rows-1, 2:columns-1), [3 3]);
blueMF = medfilt2(blueChannel(2:rows-1, 2:columns-1), [3 3]);
redTemp = redChannel; % Create a copy of red channel
redTemp(2:rows-1, 2:columns-1) = redMF; % Skip first and last column and row, change inside with redMF
redMF = redTemp; % Assign new redMF
greenTemp = greenChannel;
greenTemp(2:rows-1, 2:columns-1) = greenMF;
greenMF = greenTemp;
blueTemp = blueChannel;
blueTemp(2:rows-1, 2:columns-1) = blueMF;
blueMF = blueTemp;

This process will maintain image's size for next lines to prevent errors like you had.

2. noiseImage is a binary image with size of the main picture. It's 1 where greenChannel's value is 0 or 255.

noiseFreeGreen(noiseImage) = greenMF(noiseImage);

This line will change only operates where noiseImage pixel's value is 1. This line will replace 0 or 255 value at pixel with corresponding value from median filtered image. Instead of blurring all the picture, it will only focus on the 0 and 255 values.