comparison util/Rice Wavelet Toolbox/denoise.m @ 78:f69ae88b8be5

added Rice Wavelet Toolbox with my modification, so it can be compiled on newer systems.
author Ivan Damnjanovic lnx <ivan.damnjanovic@eecs.qmul.ac.uk>
date Fri, 25 Mar 2011 15:27:33 +0000
parents
children
comparison
equal deleted inserted replaced
76:d052ec5b742f 78:f69ae88b8be5
1 function [xd,xn,option] = denoise(x,h,type,option)
2 % [xd,xn,option] = denoise(x,h,type,option);
3 %
4 % DENOISE is a generic program for wavelet based denoising.
5 % The program will denoise the signal x using the 2-band wavelet
6 % system described by the filter h using either the traditional
7 % discrete wavelet transform (DWT) or the linear shift invariant
8 % discrete wavelet transform (also known as the undecimated DWT
9 % (UDWT)).
10 %
11 % Input:
12 % x : 1D or 2D signal to be denoised
13 % h : Scaling filter to be applied
14 % type : Type of transform (Default: type = 0)
15 % 0 --> Discrete wavelet transform (DWT)
16 % 1 --> Undecimated DWT (UDWT)
17 % option : Default settings is marked with '*':
18 % *type = 0 --> option = [0 3.0 0 0 0 0]
19 % type = 1 --> option = [0 3.6 0 1 0 0]
20 % option(1) : Whether to threshold low-pass part
21 % 0 --> Don't threshold low pass component
22 % 1 --> Threshold low pass component
23 % option(2) : Threshold multiplier, c. The threshold is
24 % computed as:
25 % thld = c*MAD(noise_estimate)).
26 % The default values are:
27 % c = 3.0 for the DWT based denoising
28 % c = 3.6 for the UDWT based denoising
29 % option(3) : Type of variance estimator
30 % 0 --> MAD (mean absolute deviation)
31 % 1 --> STD (classical numerical std estimate)
32 % option(4) : Type of thresholding
33 % 0 --> Soft thresholding
34 % 1 --> Hard thresholding
35 % option(5) : Number of levels, L, in wavelet decomposition. By
36 % setting this to the default value '0' a maximal
37 % decomposition is used.
38 % option(6) : Actual threshold to use (setting this to
39 % anything but 0 will mean that option(3)
40 % is ignored)
41 %
42 % Output:
43 % xd : Estimate of noise free signal
44 % xn : The estimated noise signal (x-xd)
45 % option : A vector of actual parameters used by the
46 % program. The vector is configured the same way as
47 % the input option vector with one added element
48 % option(7) = type.
49 %
50 % HERE'S AN EASY WAY TO RUN THE EXAMPLES:
51 % Cut-and-paste the example you want to run to a new file
52 % called ex.m, for example. Delete out the % at the beginning
53 % of each line in ex.m (Can use search-and-replace in your editor
54 % to replace it with a space). Type 'ex' in matlab and hit return.
55 %
56 % Example 1:
57 % h = daubcqf(6); [s,N] = makesig('Doppler'); n = randn(1,N);
58 % x = s + n/10; % (approximately 10dB SNR)
59 % figure;plot(x);hold on;plot(s,'r');
60 %
61 % %Denoise x with the default method based on the DWT
62 % [xd,xn,opt1] = denoise(x,h);
63 % figure;plot(xd);hold on;plot(s,'r');
64 %
65 % %Denoise x using the undecimated (LSI) wavelet transform
66 % [yd,yn,opt2] = denoise(x,h,1);
67 % figure;plot(yd);hold on;plot(s,'r');
68 %
69 % Example 2: (on an image)
70 % h = daubcqf(6); load lena;
71 % noisyLena = lena + 25 * randn(size(lena));
72 % figure; colormap(gray); imagesc(lena); title('Original Image');
73 % figure; colormap(gray); imagesc(noisyLena); title('Noisy Image');
74 % Denoise lena with the default method based on the DWT
75 % [denoisedLena,xn,opt1] = denoise(noisyLena,h);
76 % figure; colormap(gray); imagesc(denoisedLena); title('denoised Image');
77 %
78 %
79 % See also: mdwt, midwt, mrdwt, mirdwt, SoftTh, HardTh, setopt
80 %
81
82 %File Name: denoise.m
83 %Last Modification Date: 04/15/97 10:44:28
84 %Current Version: denoise.m 2.4
85 %File Creation Date: Mon Feb 20 08:33:15 1995
86 %Author: Jan Erik Odegard <odegard@ece.rice.edu>
87 %
88 %Copyright (c) 2000 RICE UNIVERSITY. All rights reserved.
89 %Created by Jan Erik Odegard, Department of ECE, Rice University.
90 %
91 %This software is distributed and licensed to you on a non-exclusive
92 %basis, free-of-charge. Redistribution and use in source and binary forms,
93 %with or without modification, are permitted provided that the following
94 %conditions are met:
95 %
96 %1. Redistribution of source code must retain the above copyright notice,
97 % this list of conditions and the following disclaimer.
98 %2. Redistribution in binary form must reproduce the above copyright notice,
99 % this list of conditions and the following disclaimer in the
100 % documentation and/or other materials provided with the distribution.
101 %3. All advertising materials mentioning features or use of this software
102 % must display the following acknowledgment: This product includes
103 % software developed by Rice University, Houston, Texas and its contributors.
104 %4. Neither the name of the University nor the names of its contributors
105 % may be used to endorse or promote products derived from this software
106 % without specific prior written permission.
107 %
108 %THIS SOFTWARE IS PROVIDED BY WILLIAM MARSH RICE UNIVERSITY, HOUSTON, TEXAS,
109 %AND CONTRIBUTORS AS IS AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING,
110 %BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
111 %FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL RICE UNIVERSITY
112 %OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
113 %EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
114 %PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS;
115 %OR BUSINESS INTERRUPTIONS) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
116 %WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
117 %OTHERWISE), PRODUCT LIABILITY, OR OTHERWISE ARISING IN ANY WAY OUT OF THE
118 %USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
119 %
120 %For information on commercial licenses, contact Rice University's Office of
121 %Technology Transfer at techtran@rice.edu or (713) 348-6173
122 %
123 %Change History: Fixed output of function and an error in the computation
124 % of the threshold for redundant denoising.
125 % <Jan Erik Odegard> <Mon Jul 31, 1995>
126 %
127 % This code is composed of several of our old codes for
128 % wavelet based denoising. In an effort to make the mess
129 % more manageable we decided to create on code that would
130 % handle all the various wavelet based denoising methods.
131 % However, only time will show (as we discover new and
132 % improved forms of denoising) if we can succeed in our goals.
133 % <Jan Erik Odegard> <Thu May 11, 1995>
134 %
135
136 if(nargin < 2)
137 error('You need to provide at least 2 inputs: x and h');
138 end;
139 if(nargin < 3),
140 type = 0;
141 option = [];
142 elseif(nargin < 4)
143 option = [];
144 end;
145 if(isempty(type)),
146 type = 0;
147 end;
148 if(type == 0),
149 default_opt = [0 3.0 0 0 0 0];
150 elseif(type == 1),
151 default_opt = [0 3.6 0 1 0 0];
152 else,
153 error(['Unknown denoising method',10,...
154 'If it is any good we need to have a serious talk :-)']);
155 end;
156 option = setopt(option,default_opt);
157 [mx,nx] = size(x);
158 dim = min(mx,nx);
159 if(dim == 1),
160 n = max(mx,nx);
161 else,
162 n = dim;
163 end;
164 if(option(5) == 0),
165 L = floor(log2(n));
166 else
167 L = option(5);
168 end;
169 if(type == 0), % Denoising by DWT
170 xd = mdwt(x,h,L);
171 if (option(6) == 0),
172 tmp = xd(floor(mx/2)+1:mx,floor(nx/2)+1:nx);
173 if(option(3) == 0),
174 thld = option(2)*median(abs(tmp(:)))/.67;
175 elseif(option(3) == 1),
176 thld = option(2)*std(tmp(:));
177 else
178 error('Unknown threshold estimator, Use either MAD or STD');
179 end;
180 else,
181 thld = option(6);
182 end;
183 if(dim == 1)
184 ix = 1:n/(2^L);
185 ykeep = xd(ix);
186 else
187 ix = 1:mx/(2^L);
188 jx = 1:nx/(2^L);
189 ykeep = xd(ix,jx);
190 end;
191 if(option(4) == 0),
192 xd = SoftTh(xd,thld);
193 elseif(option(4) == 1),
194 xd = HardTh(xd,thld);
195 else,
196 error('Unknown threshold rule. Use either Soft (0) or Hard (1)');
197 end;
198 if (option(1) == 0),
199 if(dim == 1),
200 xd(ix) = ykeep;
201 else,
202 xd(ix,jx) = ykeep;
203 end;
204 end;
205 xd = midwt(xd,h,L);
206 elseif(type == 1), % Denoising by UDWT
207 [xl,xh] = mrdwt(x,h,L);
208 if(dim == 1),
209 c_offset = 1;
210 else,
211 c_offset = 2*nx + 1;
212 end;
213 if (option(6) == 0),
214 tmp = xh(:,c_offset:c_offset+nx-1);
215 if(option(3) == 0),
216 thld = option(2)*median(abs(tmp(:)))/.67;
217 elseif(option(3) == 1),
218 thld = option(2)*std(tmp(:));
219 else
220 error('Unknown threshold estimator, Use either MAD or STD');
221 end;
222 else,
223 thld = option(6);
224 end;
225 if(option(4) == 0),
226 xh = SoftTh(xh,thld);
227 if(option(1) == 1),
228 xl = SoftTh(xl,thld);
229 end;
230 elseif(option(4) == 1),
231 xh = HardTh(xh,thld);
232 if(option(1) == 1),
233 xl = HardTh(xl,thld);
234 end;
235 else,
236 error('Unknown threshold rule. Use either Soft (0) or Hard (1)');
237 end;
238 xd = mirdwt(xl,xh,h,L);
239 else, % Denoising by unknown method
240 error(['Unknown denoising method',10,...
241 'If it is any good we need to have a serious talk :-)']);
242 end;
243 option(6) = thld;
244 option(7) = type;
245 xn = x - xd;