I want to multiply two matrices by sliding the smaller matrix on the larger one through neighborhood operations

These two matrices are a result of log-polar map and then fft2 and fft2shift has been applied.

Now I want to apply the resulting matrices to get maximum average correlation peaks.

So, there are two matrices one is THETA_F1 with size of 512x642 and THETA_F2 with size of 245x321.

When I apply THETA_F1.*THETA_F2 whenever I apply this operation I get peaks in one region.

I want to get peak throughout which is only possible if I slide THETA_F2 over THETA_F1 and correlated them.

I tried nlfilter, blockproc but it isn't seem to work. How can I slide a kernel (THETA_F2) over THETA_F1 and multiply them and then plot the result.

Answer 1

[Subhadeep Koley]

It seems the functon conv2 does what you are trying to achieve.

THETA_F1 = randi(512, 642); % Demp data
THETA_F2 = randi(245, 321); % Demo data
convolvedImg = conv2(THETA_F1, THETA_F2, 'same');
figure; imshow(convolvedImg, []);

See if this helps!

[maaham banu]

Hi, I have already tried this method, it is not helping me.