Question

I have an array of average RGB pixel values from a ROI out of 611 images (dimensions 3×611). The images are taken over 64 seconds. I interpret avergae values of each color stream as a signal with time and plot it in Matlab. I also plot the power spectrum using fft(). In this power spectrum I am getting a huge values in the low frequency ranges. I want to remove these values and so want a high pass filter with cutoff at 0.2 Hz. Also I want this filter to be very sharp and totally dampen the power of frequencies less than 0.2 Hz and maintain the power of frequencies greater than 0.2 Hz. Please suggest a way to do this.

Code till I get power spectrum of green stream

imageFiles = dir('*.png');      
noFiles = length(imageFiles);    % Number of files found
avgVals = zeros(noFiles,3);

noise = 0;                       %Boolean for noise: if 1, then ROI moves randomly

noiseV = 0;
noiseH = 0;

for i=1:noFiles
   curFileName = imageFiles(i).name;
   curImage = imread(curFileName);

   temp1  = curImage(175:215 , 290:385, 1);
   avgVals(i,1) = (sum(sum(temp1)'))/3936;

   temp2  = curImage(175:215 , 290:385, 2);
   avgVals(i,2) = (sum(sum(temp2)'))/3936;

   temp3  = curImage(175:215 , 290:385, 3);
   avgVals(i,3) = (sum(sum(temp3)'))/3936;
end

Rstream = avgVals(:,1);
Gstream = avgVals(:,2);
Bstream = avgVals(:,3);

%plot actual signal with respect to time
xAxis = (0:64/610:64);
plot(xAxis,Rstream,'r');
hold on;
plot(xAxis,Gstream,'g');
hold on;
plot(xAxis,Bstream,'b');

nfft = 1024;

R = fft(Rstream,nfft);
R = R(1:nfft/2);
mR = abs(R);
G = fft(Gstream,nfft);
G = G(1:nfft/2);
mG = abs(G);
B = fft(Bstream,nfft);
B = B(1:nfft/2);
mB = abs(B);

f=(0:nfft/2-1)*(611/64)/nfft;

figure(2);
plot(f,mG,'g');

Thanks. Any help would be appreciated.

Answer 1

If you’re not familiar with Matlab filters you can use the filterbuilder GUI by typing filterbuilder in the console.

Once you made your settings you’ll get a filter object that you can use in combination with the filter function.

http://www.mathworks.de/de/help/signal/ref/filterbuilder.html

http://www.mathworks.de/de/help/matlab/ref/filter.html

If you are working on what I think you’re doing you can also do a simple detrending using detrend, in order to get rid of the “DC” part of you signal.