To Apply 2D sliding window on data to calculate doppler parameter rapidly

I need help to merge 2-for loops using filter2 (2D) filter or any suitable way to make my code efficient enough.

I want to run a 2D sliding window of 256*256 to move through whole data and calculate one 'dc' for each patch, moreover, I just received last value of ''dc'' while rest of previous values are 0.

img= [1:1664,1:4352] %% Size of image

PRF = 1925;

k = 128; % size of patch

C= zeros(size(img));

for i = k+1:size(img_o,1)-k

for j = k+1:size(img_o,2)-k

Bb = img_o(i-kk:i+kk-1, j-kk:j+kk-1);

dc(k) = doppler_centroid(Bb, PRF); %% this function calculate doppler centroid ''dc''

C(i,j) = dc(k);

end;

end;

Answer 1

[Shashank Gupta]

Hi Amjad,

There are many things, I don't understand in the code. First line defining the variable "img" seems odd because the way you defined it give the same of "img" as [ 1 ,6016] which I don't think is intended. May be that's the reason of you getting abrubt values in C. You can simply define the size of image in img and define C accordingly

img = [1664,4352] % Size of image.
C = zeros(img);  % Define C according to the size of image.

Also, yes these can be optimized, I don't know what the doppler_centroid function do. But you can remove both the for loop and convert this problem in matrix space and then use a convolution filter to solve the problem, Refer conv2 function. I will give an example describing how it can be useful.

% Let try to find a simple mean of the sliding window.
A = rand(10,10);
B = ones(2,2)./4;    % Design a matrix for taking average.
C = conv2(A,B,"same");  % This matrix is the result of average over a sliding window of length 2.

Do try out the above code and adapt in your case.

I hope this helps.

Cheers.