Search code examples
matlabimage-processingspectrum

Calculating the spectrum magnitude in an area in Matlab

I try to FFT a set of data (they are 10 pictures, with a 'T' draw in the middle) and calculate the spectrum magnitude in given areas.

Here is my code:

for i=1:10
    filename = sprintf('T%d.GIF', i);
    %fourier transform and showing
    f = imread(filename);       %read in image
    z = fft2(double(f));        % do fourier transform
    q = fftshift(z);            % puts u=0,v=0 in the centre
    Magq = log(abs(q));              % magnitude spectrum
    Phaseq=angle(q);            % phase spectrum
    imagesc(log(abs(q)+1));     % Usually for viewing purposes: 
    colorbar;
end

The outcome after fft pic

The picture above is the outcome after running my code. And the picture below is the areas I want to calculate.

areas

I'm using 

T(i,1)=mean(mean(Magq(1:150,300:340)))+mean(mean(Magq(260:400,300:340)))+mean(mean(Magq(190:210,160:320)))+mean(mean(Magq(190:210,320:480)));

to calculate the spectrum magnitude, but the result of this is always -Inf. This is mainly because mean(mean(Magq(1:150,300:340))) and mean(mean(Magq(260:400,300:340))) are both equal to -Inf.

Could anyone please give me any hints why the result is -Inf?

Solution

  • I would guess that some values of your q variable are zero, so Magq = log(abs(q)) will return -Inf at those values. Maybe the mean that you are looking for should be

    log(abs(mean(mean(q(1:150,300:340)))))

    Otherwise, if you just want to ignore the infinities, set them all to zero

    Magq(~isfinite(Magq)) = 0

    Before applying the mean function as you did before