Extracting image patches around each pixel

4 views (last 30 days)
VIJENDRA on 6 Oct 2016
Commented: Devkumar Das on 2 Aug 2017
Suppose I have an image of size 256*256
Image = imread('cameraman.tif');
I have to extract patches of 7*7 around each pixel of image
patchSize = 7;
Around edges I padded image with zeros pixel
padcam = padarray(Image,[7 7],'both');
Total number of pixels is 256*256 therefore total patches would be 65536
[m,n] = size(padcam);
ImagePatches = zeros(patchSize,patchSize,256*256);
for k = 1:65536
for i = 8:263
for j = 8:263
ImagePatches(:,:,k) = padcam(i-3:i+3, j-3:j+3);
Due to 3 for-loops the operation is taking very long time. How can i optimize this code?
Adam on 6 Oct 2016
Edited: Adam on 6 Oct 2016
You should only need to loop twice. One loop is redundant. There are probably quicker methods that I don't have time to think about so someone else will likely suggest them, but instead of your outer k loop just keep a counter that you initialise and increment within your inner loop to give you your k index into the ImagePatches.

Sign in to comment.

Answers (1)

Guillaume on 7 Oct 2016
Since you have the imaging toolbox, a much simpler way to generate your output is by subverting nlfilter. You don't even need to pad your image as nlfilter does that for you:
ImagePatches = nlfilter(Image, [7 7], @(block) {block}); %store each block generated by nlfilter into scalar cell of cell array
ImagePatches = cat(3, ImagePatches{:}); %and concatenate in 3rd dimension
  1 Comment
Devkumar Das
Devkumar Das on 2 Aug 2017
Hi Guillaume, If I want to extract patch for a color image then nlfilter will not work. Can you suggest something for that?

Sign in to comment.


Find more on Images in Help Center and File Exchange

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!