1 | // ----------------------------------------------------------------------- |
---|
2 | // atrous_1d_reconstruct.cc: 1-dimensional wavelet reconstruction. |
---|
3 | // ----------------------------------------------------------------------- |
---|
4 | // Copyright (C) 2006, Matthew Whiting, ATNF |
---|
5 | // |
---|
6 | // This program is free software; you can redistribute it and/or modify it |
---|
7 | // under the terms of the GNU General Public License as published by the |
---|
8 | // Free Software Foundation; either version 2 of the License, or (at your |
---|
9 | // option) any later version. |
---|
10 | // |
---|
11 | // Duchamp is distributed in the hope that it will be useful, but WITHOUT |
---|
12 | // ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or |
---|
13 | // FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License |
---|
14 | // for more details. |
---|
15 | // |
---|
16 | // You should have received a copy of the GNU General Public License |
---|
17 | // along with Duchamp; if not, write to the Free Software Foundation, |
---|
18 | // Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307, USA |
---|
19 | // |
---|
20 | // Correspondence concerning Duchamp may be directed to: |
---|
21 | // Internet email: Matthew.Whiting [at] atnf.csiro.au |
---|
22 | // Postal address: Dr. Matthew Whiting |
---|
23 | // Australia Telescope National Facility, CSIRO |
---|
24 | // PO Box 76 |
---|
25 | // Epping NSW 1710 |
---|
26 | // AUSTRALIA |
---|
27 | // ----------------------------------------------------------------------- |
---|
28 | #include <iostream> |
---|
29 | #include <sstream> |
---|
30 | #include <iomanip> |
---|
31 | #include <math.h> |
---|
32 | #include <duchamp.hh> |
---|
33 | #include <param.hh> |
---|
34 | #include <Utils/utils.hh> |
---|
35 | #include <Utils/feedback.hh> |
---|
36 | #include <ATrous/atrous.hh> |
---|
37 | #include <ATrous/filter.hh> |
---|
38 | #include <Utils/Statistics.hh> |
---|
39 | using Statistics::madfmToSigma; |
---|
40 | |
---|
41 | namespace duchamp |
---|
42 | { |
---|
43 | |
---|
44 | void atrous1DReconstruct(long &xdim, float *&input, float *&output, Param &par) |
---|
45 | { |
---|
46 | /** |
---|
47 | * A routine that uses the a trous wavelet method to reconstruct a |
---|
48 | * 1-dimensional spectrum. |
---|
49 | * The Param object "par" contains all necessary info about the filter and |
---|
50 | * reconstruction parameters. |
---|
51 | * |
---|
52 | * If all pixels are BLANK (and we are testing for BLANKs), the |
---|
53 | * reconstruction will simply give BLANKs back, so we return the |
---|
54 | * input array as the output array. |
---|
55 | * |
---|
56 | * \param xdim The length of the spectrum. |
---|
57 | * \param input The input spectrum. |
---|
58 | * \param output The returned reconstructed spectrum. This array needs to |
---|
59 | * be declared beforehand. |
---|
60 | * \param par The Param set. |
---|
61 | */ |
---|
62 | |
---|
63 | const float SNR_THRESH=par.getAtrousCut(); |
---|
64 | const int MIN_SCALE=par.getMinScale(); |
---|
65 | static bool firstTime = true; |
---|
66 | |
---|
67 | int numScales = par.filter().getNumScales(xdim); |
---|
68 | int maxScale = par.getMaxScale(); |
---|
69 | if((maxScale>0)&&(maxScale<=numScales)) |
---|
70 | maxScale = std::min(maxScale,numScales); |
---|
71 | else{ |
---|
72 | if((firstTime)&&(maxScale!=0)){ |
---|
73 | firstTime=false; |
---|
74 | std::stringstream errmsg; |
---|
75 | errmsg << "The requested value of the parameter scaleMax, \"" |
---|
76 | << maxScale << "\" is outside the allowed range (1-" |
---|
77 | << numScales <<").\nSetting to " << numScales << ".\n"; |
---|
78 | duchampWarning("Reading parameters",errmsg.str()); |
---|
79 | } |
---|
80 | maxScale = numScales; |
---|
81 | } |
---|
82 | double *sigmaFactors = new double[numScales+1]; |
---|
83 | for(int i=0;i<=numScales;i++){ |
---|
84 | if(i<=par.filter().maxFactor(1)) |
---|
85 | sigmaFactors[i] = par.filter().sigmaFactor(1,i); |
---|
86 | else sigmaFactors[i] = sigmaFactors[i-1] / sqrt(2.); |
---|
87 | } |
---|
88 | |
---|
89 | float mean,sigma,originalSigma,originalMean,oldsigma,newsigma; |
---|
90 | bool *isGood = new bool[xdim]; |
---|
91 | int goodSize=0; |
---|
92 | for(int pos=0;pos<xdim;pos++) { |
---|
93 | isGood[pos] = !par.isBlank(input[pos]); |
---|
94 | if(isGood[pos]) goodSize++; |
---|
95 | } |
---|
96 | |
---|
97 | if(goodSize == 0){ |
---|
98 | // There are no good pixels -- everything is BLANK for some reason. |
---|
99 | // Return the input array as the output. |
---|
100 | |
---|
101 | for(int pos=0;pos<xdim; pos++) output[pos] = input[pos]; |
---|
102 | |
---|
103 | } |
---|
104 | else{ |
---|
105 | // Otherwise, all is good, and we continue. |
---|
106 | |
---|
107 | |
---|
108 | float *coeffs = new float[xdim]; |
---|
109 | float *wavelet = new float[xdim]; |
---|
110 | |
---|
111 | for(int pos=0;pos<xdim;pos++) output[pos]=0.; |
---|
112 | |
---|
113 | int filterHW = par.filter().width()/2; |
---|
114 | double *filter = new double[par.filter().width()]; |
---|
115 | for(int i=0;i<par.filter().width();i++) filter[i] = par.filter().coeff(i); |
---|
116 | |
---|
117 | |
---|
118 | // No trimming done in 1D case. |
---|
119 | |
---|
120 | int iteration=0; |
---|
121 | newsigma = 1.e9; |
---|
122 | do{ |
---|
123 | if(par.isVerbose()) { |
---|
124 | std::cout << "Iteration #"<<++iteration<<":"; |
---|
125 | printSpace(13); |
---|
126 | } |
---|
127 | // first, get the value of oldsigma and set it to the previous |
---|
128 | // newsigma value |
---|
129 | oldsigma = newsigma; |
---|
130 | // all other times round, we are transforming the residual array |
---|
131 | for(int i=0;i<xdim;i++) coeffs[i] = input[i] - output[i]; |
---|
132 | |
---|
133 | float *array = new float[xdim]; |
---|
134 | goodSize=0; |
---|
135 | for(int i=0;i<xdim;i++) if(isGood[i]) array[goodSize++] = input[i]; |
---|
136 | findMedianStats(array,goodSize,originalMean,originalSigma); |
---|
137 | originalSigma = madfmToSigma(originalSigma); |
---|
138 | delete [] array; |
---|
139 | |
---|
140 | int spacing = 1; |
---|
141 | for(int scale = 1; scale<=numScales; scale++){ |
---|
142 | |
---|
143 | if(par.isVerbose()) { |
---|
144 | // printBackSpace(12); |
---|
145 | std::cout << "Scale " << std::setw(2) << scale |
---|
146 | << " /" << std::setw(2) << numScales <<std::flush; |
---|
147 | } |
---|
148 | |
---|
149 | for(int xpos = 0; xpos<xdim; xpos++){ |
---|
150 | // loops over each pixel in the image |
---|
151 | int pos = xpos; |
---|
152 | |
---|
153 | wavelet[pos] = coeffs[pos]; |
---|
154 | |
---|
155 | if(!isGood[pos] ) wavelet[pos] = 0.; |
---|
156 | else{ |
---|
157 | |
---|
158 | for(int xoffset=-filterHW; xoffset<=filterHW; xoffset++){ |
---|
159 | int x = xpos + spacing*xoffset; |
---|
160 | |
---|
161 | while((x<0)||(x>=xdim)){ |
---|
162 | // boundary conditions are reflection. |
---|
163 | if(x<0) x = 0 - x; |
---|
164 | else if(x>=xdim) x = 2*(xdim-1) - x; |
---|
165 | } |
---|
166 | |
---|
167 | int filterpos = (xoffset+filterHW); |
---|
168 | int oldpos = x; |
---|
169 | |
---|
170 | if(isGood[oldpos]) |
---|
171 | wavelet[pos] -= filter[filterpos]*coeffs[oldpos]; |
---|
172 | |
---|
173 | } //-> end of xoffset loop |
---|
174 | } //-> end of else{ ( from if(!isGood[pos]) ) |
---|
175 | |
---|
176 | } //-> end of xpos loop |
---|
177 | |
---|
178 | // Need to do this after we've done *all* the convolving |
---|
179 | for(int pos=0;pos<xdim;pos++) coeffs[pos] = coeffs[pos] - wavelet[pos]; |
---|
180 | |
---|
181 | // Have found wavelet coeffs for this scale -- now threshold |
---|
182 | if(scale>=MIN_SCALE){ |
---|
183 | array = new float[xdim]; |
---|
184 | goodSize=0; |
---|
185 | for(int pos=0;pos<xdim;pos++) |
---|
186 | if(isGood[pos]) array[goodSize++] = wavelet[pos]; |
---|
187 | findMedianStats(array,goodSize,mean,sigma); |
---|
188 | delete [] array; |
---|
189 | |
---|
190 | for(int pos=0;pos<xdim;pos++){ |
---|
191 | // preserve the Blank pixel values in the output. |
---|
192 | if(!isGood[pos]) output[pos] = input[pos]; |
---|
193 | else if( fabs(wavelet[pos]) > |
---|
194 | (mean+SNR_THRESH*originalSigma*sigmaFactors[scale]) ) |
---|
195 | output[pos] += wavelet[pos]; |
---|
196 | } |
---|
197 | } |
---|
198 | |
---|
199 | spacing *= 2; |
---|
200 | |
---|
201 | } //-> end of scale loop |
---|
202 | |
---|
203 | // Only add the final smoothed array if we are doing *all* the scales. |
---|
204 | if(numScales == par.filter().getNumScales(xdim)) |
---|
205 | for(int pos=0;pos<xdim;pos++) |
---|
206 | if(isGood[pos]) output[pos] += coeffs[pos]; |
---|
207 | |
---|
208 | array = new float[xdim]; |
---|
209 | goodSize=0; |
---|
210 | for(int i=0;i<xdim;i++) |
---|
211 | if(isGood[i]) array[goodSize++] = input[i] - output[i]; |
---|
212 | findMedianStats(array,goodSize,mean,newsigma); |
---|
213 | newsigma = madfmToSigma(newsigma); |
---|
214 | delete [] array; |
---|
215 | |
---|
216 | if(par.isVerbose()) printBackSpace(26); |
---|
217 | |
---|
218 | } while( (iteration==1) || |
---|
219 | (fabs(oldsigma-newsigma)/newsigma > reconTolerance) ); |
---|
220 | |
---|
221 | if(par.isVerbose()) std::cout << "Completed "<<iteration<<" iterations. "; |
---|
222 | |
---|
223 | delete [] filter; |
---|
224 | delete [] coeffs; |
---|
225 | delete [] wavelet; |
---|
226 | |
---|
227 | } |
---|
228 | |
---|
229 | delete [] isGood; |
---|
230 | delete [] sigmaFactors; |
---|
231 | } |
---|
232 | |
---|
233 | } |
---|