How to find out windowwise mean and standard deviation of some particular pixels.

I have a .txt file with 50 rows and 2 columns indicating 50 pixel positions in an image.I am reading that file with the following statement:

file_id=fopen('C:\Users\Mahua Nandy\PixelPositions_im1.txt','r'); Input=fscanf(file_id,'%d',[2,50]); fclose(file_id);

Input is coming in column order(2 rows and 50 columns).Now I want to find out 3x3 windowwise mean and standard deviation of those particular 50 pixels.

Please help me to code it

Answer 1

[Image Analyst]

First I guess you have to create an image - all black I guess.

maxRows = max(Input(1,:));
maxColumns = max(Input(2, :));
grayImage = zeros(maxrows, maxColumns, 'uint8');

Then you have to run down your list of coordinates, figuring out their gray level (from some other image I assume), then assigning/transferring them to the new image (the one that contains only the 50 pixels):

for c = 1 : maxColumns
  for r = 1 : maxRows
    row = Input(1, r);
    column = Input(2, c);
    % Transfer the gray level from the other image to
    % this location on the new image.
    grayImage(row, column) = initialImage(row, column);
  end
end

Then you have to use conv2() and stdfilt() to take the mean and std dev in a 3x3 window in the new image which contains the 50 pixels.

meanImage = conv2(grayImage, ones(3)/9, 'same');
stdImage = stdfilt(grayImage);

Be aware though that most of the image will be just black pixels so the mean will be very low. Upload your image first, if you have any follow up questions.

[Image Analyst]

Well, did this work for you? Are you going to respond?

[Mahua Nandy(Pal)]

Thank you vry much for the reply.But still I can not figure out my solutions.Please help. Following are the 50 different pixels of my image.I want to find out mean and std deviation of each pixel considering its 3x3 neighborhood in the original image.

28 16
28 17
28 18
28 27
28 28
30 57
30 58
30 59
30 60
30 61
31 56
31 57
31 58
31 59
31 60
31 115
31 116
31 117
31 118
31 119
32 39
32 40
32 41
32 42
32 43
33 28
33 29
33 30
33 31
33 32
33 118
33 119
33 120
33 121
33 122
34 39
34 40
34 41
34 42
34 43
35 102
35 103
35 104
35 105
35 106
37 74
37 75
37 76
37 77
37 78
38 38
38 39
38 40
38 41
38 42
111 131
111 132
111 133
111 134
111 135

[Image Analyst]

% Gets mean and std dev of an image around 50 isolated coordinates.
clc;    % Clear the command window.
close all;  % Close all figures (except those of imtool.)
imtool close all;  % Close all imtool figures.
clear all;  % Erase all existing variables.
workspace;  % Make sure the workspace panel is showing.
fontSize = 20;
format compact;

Input = [...
  28 16
  28 17
  28 18
  28 27
  28 28
  30 57
  30 58
  30 59
  30 60
  30 61
  31 56
  31 57
  31 58
  31 59
  31 60
  31 115
  31 116
  31 117
  31 118
  31 119
  32 39
  32 40
  32 41
  32 42
  32 43
  33 28
  33 29
  33 30
  33 31
  33 32
  33 118
  33 119
  33 120
  33 121
  33 122
  34 39
  34 40
  34 41
  34 42
  34 43
  35 102
  35 103
  35 104
  35 105
  35 106
  37 74
  37 75
  37 76
  37 77
  37 78
  38 38
  38 39
  38 40
  38 41
  38 42
  111 131
  111 132
  111 133
  111 134
  111 135]

initialImage = imread('moon.tif');

maxRows = max(Input(:, 2)); % Y value
maxColumns = max(Input(:, 1)); % x value.
grayImage = zeros(maxRows, maxColumns, 'uint8');
numberOfCoordinates = size(Input, 1);
for c = 1 : numberOfCoordinates
  for r = 1 : numberOfCoordinates
    row = Input(r, 1);
    column = Input(c, 2);
    % Transfer the gray level from the other image to
    % this location on the new image.
    grayImage(row, column) = initialImage(row, column);
  end
end

meanImage = conv2(double(grayImage), ones(3)/9, 'same');
subplot(1, 2, 1);
imshow(meanImage, []);
title('Mean Image', 'FontSize', fontSize);
stdImage = stdfilt(grayImage);
subplot(1, 2, 2);
imshow(stdImage, []);
title('Std Dev Image', 'FontSize', fontSize);
% Enlarge figure to full screen.
set(gcf, 'units','normalized','outerposition',[0 0 1 1]);

[Mahua Nandy(Pal)]

thank you very much but I need 3x3 windowwise mean and standard deviation of those particular pixels where the 3x3 window is the window of original image not new image.Thank you once again for the answer because I think the rest part I can manage.

[Image Analyst]

That's what conv2 does. And what exactly is in the space between your 50 isolated pixels? In an image, there has to be SOMETHING.

[Mahua Nandy(Pal)]

thank you

Answer 2

[Thorsten]

Hi Mahua Nandy(Pal), that's all quite straight forward:

1. Let's say you have your image in I and your particular positions in variables x and y

 I = imread('cameraman.tif'); 
 x = 10; y = 20; % just for testing

To get the 3 x 3 neighborhood N around (x, y) from your image, use

 N = I(y-1:y+1, x-1:x+1);

That gives you a 3 x 3 matrix. The mean and std are

 m = mean(N(:));
 s = std(N(:));

If you have 50 values of x and y stored in Input, you just do it in a loop:

 for i=1:size(Input, 1)

   x = Input(i, 1);
   y = Input(i, 2);
   N = I(y-1:y+1, x-1:x+1);
   m(i) =  mean(N(:));
   s(i) = std(N(:));
 end

Done!

Two caveats:

1. verify that each row of Input is [x y] and not [y x]; if the columns are given as y x use

x(i) = Input(i, 2); 
y(i) = Input(i, 1);

2. ensure that no x, y lies at the boundary of I, such that x-1, x+1m, y-1 and y+1 is always a valid index to the image. To be one the save side, you could use the statement

N = I(max(1, y-1):min(size(I, 1), y+1), ...
      max(1, x-1):min(size(I, 2), x+1);

[Mahua Nandy(Pal)]

thanks