A First Course in Wavelets with Fourier Analysis, 2nd Edition

APPENDIX C

MATLAB ROUTINES

C.1 GENERAL COMPRESSION ROUTINE

The following Matlab function compresses a given vector, thereby zeroing out the terms falling below a user specified (percentage) threshold. This routine is used by all fft and wavelet compression schemes that follow.

function wc=compress(w,r)
% Input is the array w and r, which is a number
% between 0 and 1.
% Output is the array wc where smallest 100r% of the
% terms in w are set to zero (e.g r=.75 means the smallest 75% of
% the terms are set to zero
if (r<0) | (r>1)
  error(’r should be between 0 and 1’) end;
N=length(w); Nr=floor(N*r); ww=sort(abs(w)); tol=abs(ww(Nr)); for j=1:N
  if (abs(w(j)) < tol)
w(j)=0;
  end;
end;
wc=w;

C.2 USE OF MATLAB’S FFT ROUTINE FOR FILTERING AND COMPRESSION

Filtering with FFT. The following Matlab commands are needed for Example 3.6, which discusses filtering the high-frequency components from a given signal. The key Matlab commands are fft and ifft, which compute the fast Fourier transform and its inverse, respectively.

 >> t=linspace(0,2*pi,2^8); % discretizes [0, 2pi] into 256 nodes >> y=exp(-(cos(t).^2)).*(sin(2*t)+2*cos(4*t)+0.4*sin(t).*sin(50*t)); >> plot(t,y) % generates the graph of the original signal >> fy=fft(y); % computes fft of y >> filterfy=[fy(1:6) zeros(1,2^8-12) fy(2^8-5:2^8)]; % sets fft coefficients >>% to zero for $7 \leq k \leq 256$ >> filtery=ifft(filterfy); % computes inverse fft of the filtered fft >> plot(t, filtery) % generates the plot of the compressed signal ...