mefig214.m

老外写的小波变换的工具箱

% mefig214 -- Segmented vs. Ordinary de-noising; object "Cusp" 
%
% In this display we compare segmented de-noising of figure 2.13
% with traditional de-noising using periodized nearly-symmetric wavelets
% having 8 vanishing moments.
%
% De-noising in a non-segmented wavelet transform domain leads to
% "pulling down" of peak values.  Such distortion can be avoided 
% avoided by de-noising with ideally-selected segmentation point.
%
% Note -- this script uses results calculated in mefig213.m
%



%



	global yCusp Cusp SmoothCusp
	global L
    if isempty(yCusp)
        warndlg('Please Run figure 2.13 first and then run this figure');
        return;
    end
    
 %
	qmf = MakeONFilter('Symmlet',8);
	wyCusp    = FWT_PO(yCusp,L,qmf);
	shCusp    = MultiVisu(wyCusp,L); 
	sbCusp    = IWT_PO(shCusp,L,qmf);
%
	%clf;
    	absc  = (1:length(Cusp)) ./ length(Cusp);
	
	object = 5*Cusp.^8;
    
    %subplot(111)
	plot(absc, object,':'); hold on;
	plot(absc,sbCusp, '-.'); 
	plot(absc,SmoothCusp,'-');  hold off
	title('2.14 Object Cusp (..) Unsegmented (-.) and Segmented (-) DeNoising')
	sum((object - SmoothCusp).^2)
	sum((object - sbCusp).^2)
%  
% Prepared for the paper Minimum Entropy Segmentation 
% Copyright (c) 1994 David L. Donoho
%  
    
    
 
 
%
%  Part of Wavelab Version 850
%  Built Tue Jan  3 13:20:41 EST 2006
%  This is Copyrighted Material
%  For Copying permissions see COPYING.m
%  Comments? e-mail [e-mail]