[3] | 1 | #include <iostream> |
---|
| 2 | #include <math.h> |
---|
| 3 | #include <ATrous/atrous.hh> |
---|
| 4 | //#include <Cubes/cubes.hh> |
---|
| 5 | #include <Utils/utils.hh> |
---|
| 6 | |
---|
| 7 | using namespace std; |
---|
| 8 | void atrous1DReconstructOLD(long &size, float *input,float *output, Param &par) |
---|
| 9 | { |
---|
| 10 | const float SNR_THRESH=par.getAtrousCut(); |
---|
| 11 | const int MIN_SCALE=par.getMinScale(); |
---|
| 12 | |
---|
| 13 | int numScales = getNumScales(size); |
---|
| 14 | /* |
---|
| 15 | if(numScales>maxNumScales1D){ |
---|
| 16 | cerr<<"Error in atrous1DReconstruct:: numScales ("<<numScales<<") > "<<maxNumScales1D<<"\n"; |
---|
| 17 | cerr<<"Don't have correction factors for this many scales...\n"; |
---|
| 18 | cerr<<"Exiting...\n"; |
---|
| 19 | exit(1); |
---|
| 20 | } |
---|
| 21 | */ |
---|
| 22 | double *sigmaFactors = new double[numScales+1]; |
---|
| 23 | for(int i=0;i<=numScales;i++){ |
---|
| 24 | if(i<=maxNumScales1D) sigmaFactors[i] = sigmaFactors1D[i]; |
---|
| 25 | else sigmaFactors[i] = sigmaFactors[i-1] / sqrt(2.); |
---|
| 26 | } |
---|
| 27 | |
---|
| 28 | |
---|
| 29 | double *coeffs = new double[(numScales+1)*size]; |
---|
| 30 | double *wavelet = new double[(numScales+1)*size]; |
---|
| 31 | |
---|
| 32 | atrousTransform(size,numScales,input,coeffs,wavelet); |
---|
| 33 | |
---|
| 34 | for(int pos=0;pos<size;pos++) output[pos]=0.; |
---|
| 35 | |
---|
| 36 | float *array = new float[size]; |
---|
| 37 | float mean,sigma,originalSigma,originalMean; |
---|
| 38 | findMedianStats(input,size,originalMean,originalSigma); |
---|
| 39 | for(int scale=MIN_SCALE;scale<numScales;scale++){ |
---|
| 40 | for(int pos=0;pos<size;pos++) array[pos] = wavelet[scale*size + pos]; |
---|
| 41 | findMedianStats(array,size,mean,sigma); |
---|
| 42 | for(int pos=0;pos<size;pos++){ |
---|
| 43 | if( fabs(wavelet[scale*size+pos])>(mean+SNR_THRESH*originalSigma*sigmaFactors[scale])) |
---|
| 44 | output[pos] += wavelet[scale*size+pos]; |
---|
| 45 | } |
---|
| 46 | } |
---|
| 47 | for(int pos=0;pos<size;pos++) |
---|
| 48 | output[pos] += coeffs[numScales*size+pos]; |
---|
| 49 | float *residual = new float[size]; |
---|
| 50 | for(int pos=0;pos<size;pos++) |
---|
| 51 | residual[pos] = input[pos] - output[pos]; |
---|
| 52 | float oldsigma,newsigma; |
---|
| 53 | do{ |
---|
| 54 | findMedianStats(residual,size,mean,oldsigma); |
---|
| 55 | atrousTransform(size,numScales,residual,coeffs,wavelet); |
---|
| 56 | for(int scale=MIN_SCALE;scale<numScales;scale++){ |
---|
| 57 | for(int pos=0;pos<size;pos++) array[pos] = wavelet[scale*size+pos]; |
---|
| 58 | findMedianStats(array,size,mean,sigma); |
---|
| 59 | for(int pos=0;pos<size;pos++){ |
---|
| 60 | if(fabs(wavelet[scale*size+pos]) > |
---|
| 61 | (mean+SNR_THRESH*originalSigma*sigmaFactors[scale])) |
---|
| 62 | output[pos] += wavelet[scale*size+pos]; |
---|
| 63 | } |
---|
| 64 | } |
---|
| 65 | for(int pos=0;pos<size;pos++) output[pos] += coeffs[numScales*size+pos]; |
---|
| 66 | for(int pos=0;pos<size;pos++) residual[pos] = input[pos] - output[pos]; |
---|
| 67 | findMedianStats(residual,size,mean,newsigma); |
---|
| 68 | }while( fabsf(oldsigma-newsigma)/newsigma > reconTolerance); |
---|
| 69 | |
---|
| 70 | delete [] coeffs; |
---|
| 71 | delete [] wavelet; |
---|
| 72 | delete [] residual; |
---|
| 73 | delete [] array; |
---|
| 74 | } |
---|
| 75 | |
---|
| 76 | |
---|
| 77 | |
---|
| 78 | void atrous2DReconstructOLD(long &xdim, long &ydim, float *input,float *output, Param &par) |
---|
| 79 | { |
---|
| 80 | const float SNR_THRESH=par.getAtrousCut(); |
---|
| 81 | const int MIN_SCALE=par.getMinScale(); |
---|
| 82 | bool flagBlank=par.getFlagBlankPix(); |
---|
| 83 | float blankPixValue = par.getBlankPixVal(); |
---|
| 84 | long size = xdim * ydim; |
---|
| 85 | long mindim = xdim; |
---|
| 86 | if (ydim<mindim) mindim = ydim; |
---|
| 87 | int numScales = getNumScales(mindim); |
---|
| 88 | if(numScales>maxNumScales2D){ |
---|
| 89 | cerr<<"Error in atrous2DReconstruct:: numScales ("<<numScales<<") > "<<maxNumScales2D<<"\n"; |
---|
| 90 | cerr<<"Don't have correction factors for this many scales...\n"; |
---|
| 91 | cerr<<"XDIM = "<<xdim<<", YDIM = "<<ydim<<endl; |
---|
| 92 | cerr<<"Exiting...\n"; |
---|
| 93 | exit(1); |
---|
| 94 | } |
---|
| 95 | |
---|
| 96 | double *coeffs = new double[size]; |
---|
| 97 | double *wavelet = new double[(numScales+1)*size]; |
---|
| 98 | |
---|
| 99 | atrousTransform2D(xdim,ydim,numScales,input,coeffs,wavelet,par); |
---|
| 100 | |
---|
| 101 | for(int pos=0;pos<size;pos++) output[pos]=0.; |
---|
| 102 | |
---|
| 103 | bool *isGood = new bool[size]; |
---|
| 104 | for(int pos=0;pos<size;pos++)// isGood[pos] = (!flagBlank) || (input[pos]!=blankPixValue); |
---|
| 105 | isGood[pos] = !par.isBlank(input[pos]); |
---|
| 106 | |
---|
| 107 | float mean,sigma,originalSigma,originalMean; |
---|
| 108 | // Only get stats for the non-blank pixels. |
---|
| 109 | float *array = new float[size]; |
---|
| 110 | int goodSize=0; |
---|
| 111 | for(int i=0;i<size;i++) if(isGood[i]) array[goodSize++] = input[i]; |
---|
| 112 | findMedianStats(array,goodSize,originalMean,originalSigma); |
---|
| 113 | delete [] array; |
---|
| 114 | |
---|
| 115 | for(int scale=MIN_SCALE;scale<=numScales;scale++){ |
---|
| 116 | array = new float[size]; |
---|
| 117 | goodSize=0; |
---|
| 118 | for(int pos=0;pos<size;pos++) |
---|
| 119 | if(isGood[pos]) array[goodSize++] = wavelet[scale*size + pos]; |
---|
| 120 | findMedianStats(array,goodSize,mean,sigma); |
---|
| 121 | delete [] array; |
---|
| 122 | |
---|
| 123 | for(int pos=0;pos<size;pos++){ |
---|
| 124 | // preserve the Blank pixel values in the output. |
---|
| 125 | if(!isGood[pos]) output[pos] = blankPixValue; |
---|
| 126 | else if( fabs(wavelet[scale*size+pos]) > |
---|
| 127 | (mean+SNR_THRESH*originalSigma*sigmaFactors2D[scale]) ){ |
---|
| 128 | output[pos] += wavelet[scale*size+pos]; |
---|
| 129 | } |
---|
| 130 | } |
---|
| 131 | } |
---|
| 132 | for(int pos=0;pos<size;pos++) if(isGood[pos]) output[pos] += coeffs[pos]; |
---|
| 133 | |
---|
| 134 | float *residual = new float[size]; |
---|
| 135 | for(int pos=0;pos<size;pos++) |
---|
| 136 | residual[pos] = input[pos] - output[pos]; |
---|
| 137 | |
---|
| 138 | float oldsigma,newsigma; |
---|
| 139 | do{ |
---|
| 140 | array = new float[size]; |
---|
| 141 | goodSize=0; |
---|
| 142 | for(int i=0;i<size;i++) if(isGood[i]) array[goodSize++] = residual[i]; |
---|
| 143 | findMedianStats(array,goodSize,mean,oldsigma); |
---|
| 144 | delete [] array; |
---|
| 145 | |
---|
| 146 | cerr<<"\nIn atrous2DReconstruct, setting bad bits to BLANK in residual before transform.\n"; |
---|
| 147 | for(int i=0;i<size;i++) if(!isGood[i]) residual[i]=blankPixValue; |
---|
| 148 | atrousTransform2D(xdim,ydim,numScales,residual,coeffs,wavelet,par); |
---|
| 149 | for(int i=0;i<size;i++) if(!isGood[i]) residual[i]=0; |
---|
| 150 | |
---|
| 151 | for(int scale=MIN_SCALE;scale<=numScales;scale++){ |
---|
| 152 | |
---|
| 153 | array = new float[size]; |
---|
| 154 | goodSize=0; |
---|
| 155 | for(int pos=0;pos<size;pos++) |
---|
| 156 | if(isGood[pos]) array[goodSize++] = wavelet[scale*size+pos]; |
---|
| 157 | findMedianStats(array,goodSize,mean,sigma); |
---|
| 158 | delete [] array; |
---|
| 159 | |
---|
| 160 | for(int pos=0;pos<size;pos++){ |
---|
| 161 | if(!isGood[pos]) output[pos] = blankPixValue; |
---|
| 162 | else if(fabs(wavelet[scale*size+pos]) > |
---|
| 163 | (mean+SNR_THRESH*originalSigma*sigmaFactors2D[scale])) |
---|
| 164 | output[pos] += wavelet[scale*size+pos]; |
---|
| 165 | } |
---|
| 166 | } |
---|
| 167 | for(int pos=0;pos<size;pos++) if(isGood[pos]) output[pos] += coeffs[pos]; |
---|
| 168 | for(int pos=0;pos<size;pos++) residual[pos] = input[pos] - output[pos]; |
---|
| 169 | |
---|
| 170 | array = new float[size]; |
---|
| 171 | goodSize=0; |
---|
| 172 | for(int i=0;i<size;i++) if(isGood[i]) array[goodSize++] = residual[i]; |
---|
| 173 | findMedianStats(array,goodSize,mean,newsigma); |
---|
| 174 | delete [] array; |
---|
| 175 | }while( fabsf(oldsigma-newsigma)/newsigma > reconTolerance); |
---|
| 176 | |
---|
| 177 | |
---|
| 178 | delete [] coeffs; |
---|
| 179 | delete [] wavelet; |
---|
| 180 | delete [] residual; |
---|
| 181 | } |
---|
| 182 | |
---|
| 183 | |
---|
| 184 | void atrous3DReconstructOLD(long &xdim, long &ydim, long &zdim, float *&input,float *&output, Param &par) |
---|
| 185 | { |
---|
| 186 | const float SNR_THRESH=par.getAtrousCut(); |
---|
| 187 | const int MIN_SCALE=par.getMinScale(); |
---|
| 188 | bool flagBlank=par.getFlagBlankPix(); |
---|
| 189 | float blankPixValue = par.getBlankPixVal(); |
---|
| 190 | long size = xdim * ydim * zdim; |
---|
| 191 | long mindim = xdim; |
---|
| 192 | if (ydim<mindim) mindim = ydim; |
---|
| 193 | if (zdim<mindim) mindim = zdim; |
---|
| 194 | int numScales = getNumScales(mindim); |
---|
| 195 | if(numScales>maxNumScales3D){ |
---|
| 196 | cerr<<"Error in atrous3DReconstruct:: numScales ("<<numScales<<") > "<<maxNumScales3D<<"\n"; |
---|
| 197 | cerr<<"Don't have correction factors for this many scales...\n"; |
---|
| 198 | cerr<<"Exiting...\n"; |
---|
| 199 | exit(1); |
---|
| 200 | } |
---|
| 201 | |
---|
| 202 | float mean,sigma,originalSigma,originalMean; |
---|
| 203 | bool *isGood = new bool[size]; |
---|
| 204 | for(int pos=0;pos<size;pos++)// isGood[pos] = (!flagBlank) || (input[pos]!=blankPixValue); |
---|
| 205 | isGood[pos] = !par.isBlank(input[pos]); |
---|
| 206 | |
---|
| 207 | float *array = new float[size]; |
---|
| 208 | int goodSize=0; |
---|
| 209 | for(int i=0;i<size;i++) if(isGood[i]) array[goodSize++] = input[i]; |
---|
| 210 | findMedianStats(array,goodSize,originalMean,originalSigma); |
---|
| 211 | delete [] array; |
---|
| 212 | |
---|
| 213 | float *coeffs = new float[size]; |
---|
| 214 | float *wavelet = new float[(numScales+1)*size]; |
---|
| 215 | float *residual = new float[size]; |
---|
| 216 | |
---|
| 217 | cerr << size<<" "; |
---|
| 218 | atrousTransform3D(xdim,ydim,zdim,numScales,input,coeffs,wavelet,par); |
---|
| 219 | cerr <<","; |
---|
| 220 | for(int pos=0;pos<size;pos++) output[pos]=0.; |
---|
| 221 | cerr <<","; |
---|
| 222 | |
---|
| 223 | for(int scale=MIN_SCALE;scale<=numScales;scale++){ |
---|
| 224 | cerr<<","; |
---|
| 225 | array = new float[size]; |
---|
| 226 | goodSize=0; |
---|
| 227 | for(int pos=0;pos<size;pos++) |
---|
| 228 | if(isGood[pos]) array[goodSize++] = wavelet[scale*size + pos]; |
---|
| 229 | findMedianStats(array,goodSize,mean,sigma); |
---|
| 230 | delete [] array; |
---|
| 231 | |
---|
| 232 | for(int pos=0;pos<size;pos++){ |
---|
| 233 | // preserve the Blank pixel values in the output. |
---|
| 234 | if(!isGood[pos]) output[pos] = blankPixValue; |
---|
| 235 | else if( fabs(wavelet[scale*size+pos])> |
---|
| 236 | (mean+SNR_THRESH*originalSigma*sigmaFactors3D[scale])) |
---|
| 237 | output[pos] += wavelet[scale*size+pos]; |
---|
| 238 | } |
---|
| 239 | } |
---|
| 240 | for(int pos=0;pos<size;pos++) if(isGood[pos]) output[pos] += coeffs[pos]; |
---|
| 241 | |
---|
| 242 | for(int pos=0;pos<size;pos++) residual[pos] = input[pos] - output[pos]; |
---|
| 243 | |
---|
| 244 | float oldsigma,newsigma; |
---|
| 245 | cerr<<"!"; |
---|
| 246 | array = new float[size]; |
---|
| 247 | goodSize=0; |
---|
| 248 | for(int i=0;i<size;i++) if(isGood[i]) array[goodSize++] = residual[i]; |
---|
| 249 | findMedianStats(array,goodSize,mean,newsigma); |
---|
| 250 | delete [] array; |
---|
| 251 | do{ |
---|
| 252 | oldsigma = newsigma; |
---|
| 253 | cerr<<"!"<<oldsigma; |
---|
| 254 | atrousTransform3D(xdim,ydim,zdim,numScales,residual,coeffs,wavelet,par); |
---|
| 255 | cerr<<"!"; |
---|
| 256 | for(int scale=MIN_SCALE;scale<numScales;scale++){ |
---|
| 257 | |
---|
| 258 | array = new float[size]; |
---|
| 259 | goodSize=0; |
---|
| 260 | for(int pos=0;pos<size;pos++) |
---|
| 261 | if(isGood[pos]) array[goodSize++] = wavelet[scale*size+pos]; |
---|
| 262 | findMedianStats(array,goodSize,mean,sigma); |
---|
| 263 | delete [] array; |
---|
| 264 | |
---|
| 265 | for(int pos=0;pos<size;pos++){ |
---|
| 266 | if(!isGood[pos]) output[pos] = blankPixValue; |
---|
| 267 | else if( fabs(wavelet[scale*size+pos]) > |
---|
| 268 | (mean+SNR_THRESH*originalSigma*sigmaFactors3D[scale])) |
---|
| 269 | output[pos] += wavelet[scale*size+pos]; |
---|
| 270 | } |
---|
| 271 | } |
---|
| 272 | for(int pos=0;pos<size;pos++) if(isGood[pos]) output[pos] += coeffs[pos]; |
---|
| 273 | for(int pos=0;pos<size;pos++) residual[pos] = input[pos] - output[pos]; |
---|
| 274 | array = new float[size]; |
---|
| 275 | goodSize=0; |
---|
| 276 | for(int i=0;i<size;i++) if(isGood[i]) array[goodSize++] = residual[i]; |
---|
| 277 | findMedianStats(array,goodSize,mean,newsigma); |
---|
| 278 | delete [] array; |
---|
| 279 | cerr<<"|"<<newsigma<<"|"<<fabsf(oldsigma-newsigma)/newsigma; |
---|
| 280 | |
---|
| 281 | }while( fabsf(oldsigma-newsigma)/newsigma > reconTolerance); |
---|
| 282 | |
---|
| 283 | delete [] coeffs; |
---|
| 284 | delete [] wavelet; |
---|
| 285 | delete [] residual; |
---|
| 286 | } |
---|