Coverage for python/lsst/ip/diffim/makeKernelBasisList.py: 4%

108 statements  

« prev     ^ index     » next       coverage.py v7.1.0, created at 2023-02-05 18:32 -0800

1# 

2# LSST Data Management System 

3# Copyright 2008-2016 LSST Corporation. 

4# 

5# This product includes software developed by the 

6# LSST Project (http://www.lsst.org/). 

7# 

8# This program is free software: you can redistribute it and/or modify 

9# it under the terms of the GNU General Public License as published by 

10# the Free Software Foundation, either version 3 of the License, or 

11# (at your option) any later version. 

12# 

13# This program is distributed in the hope that it will be useful, 

14# but WITHOUT ANY WARRANTY; without even the implied warranty of 

15# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 

16# GNU General Public License for more details. 

17# 

18# You should have received a copy of the LSST License Statement and 

19# the GNU General Public License along with this program. If not, 

20# see <http://www.lsstcorp.org/LegalNotices/>. 

21# 

22 

23__all__ = ["makeKernelBasisList", "generateAlardLuptonBasisList"] 

24 

25from . import diffimLib 

26from lsst.log import Log 

27import numpy as np 

28 

29sigma2fwhm = 2. * np.sqrt(2. * np.log(2.)) 

30 

31 

32def makeKernelBasisList(config, targetFwhmPix=None, referenceFwhmPix=None, 

33 basisDegGauss=None, basisSigmaGauss=None, metadata=None): 

34 """Generate the delta function or Alard-Lupton kernel bases depending on the Config. 

35 Wrapper to call either `lsst.ip.diffim.makeDeltaFunctionBasisList` or 

36 `lsst.ip.diffim.generateAlardLuptonBasisList`. 

37 

38 Parameters 

39 ---------- 

40 config : `lsst.ip.diffim.PsfMatchConfigAL` 

41 Configuration object. 

42 targetFwhmPix : `float`, optional 

43 Passed on to `lsst.ip.diffim.generateAlardLuptonBasisList`. 

44 Not used for delta function basis sets. 

45 referenceFwhmPix : `float`, optional 

46 Passed on to `lsst.ip.diffim.generateAlardLuptonBasisList`. 

47 Not used for delta function basis sets. 

48 basisDegGauss : `list` of `int`, optional 

49 Passed on to `lsst.ip.diffim.generateAlardLuptonBasisList`. 

50 Not used for delta function basis sets. 

51 basisSigmaGauss : `list` of `int`, optional 

52 Passed on to `lsst.ip.diffim.generateAlardLuptonBasisList`. 

53 Not used for delta function basis sets. 

54 metadata : `lsst.daf.base.PropertySet`, optional 

55 Passed on to `lsst.ip.diffim.generateAlardLuptonBasisList`. 

56 Not used for delta function basis sets. 

57 

58 Returns 

59 ------- 

60 basisList: `list` of `lsst.afw.math.kernel.FixedKernel` 

61 List of basis kernels. 

62 

63 Notes 

64 ----- 

65 See `lsst.ip.diffim.generateAlardLuptonBasisList` and 

66 `lsst.ip.diffim.makeDeltaFunctionBasisList` for more information. 

67 

68 Raises 

69 ------ 

70 ValueError 

71 If ``config.kernelBasisSet`` has an invalid value (not "alard-lupton" or "delta-function"). 

72 """ 

73 if config.kernelBasisSet == "alard-lupton": 

74 return generateAlardLuptonBasisList(config, targetFwhmPix=targetFwhmPix, 

75 referenceFwhmPix=referenceFwhmPix, 

76 basisDegGauss=basisDegGauss, 

77 basisSigmaGauss=basisSigmaGauss, 

78 metadata=metadata) 

79 elif config.kernelBasisSet == "delta-function": 

80 kernelSize = config.kernelSize 

81 return diffimLib.makeDeltaFunctionBasisList(kernelSize, kernelSize) 

82 else: 

83 raise ValueError("Cannot generate %s basis set" % (config.kernelBasisSet)) 

84 

85 

86def generateAlardLuptonBasisList(config, targetFwhmPix=None, referenceFwhmPix=None, 

87 basisDegGauss=None, basisSigmaGauss=None, metadata=None): 

88 """Generate an Alard-Lupton kernel basis list based upon the Config and 

89 the input FWHM of the science and template images. 

90 

91 Parameters 

92 ---------- 

93 config : `lsst.ip.diffim.PsfMatchConfigAL` 

94 Configuration object for the Alard-Lupton algorithm. 

95 targetFwhmPix : `float`, optional 

96 Fwhm width (pixel) of the template exposure characteristic psf. 

97 This is the _target_ that will be matched to the science exposure. 

98 referenceFwhmPix : `float`, optional 

99 Fwhm width (pixel) of the science exposure characteristic psf. 

100 basisDegGauss : `list` of `int`, optional 

101 Polynomial degree of each Gaussian (sigma) basis. If None, defaults to `config.alardDegGauss`. 

102 basisSigmaGauss : `list` of `int`, optional 

103 Sigmas of each Gaussian basis. If None, defaults to `config.alardSigGauss`. 

104 metadata : `lsst.daf.base.PropertySet`, optional 

105 If specified, object to collect metadata fields about the kernel basis list. 

106 

107 Returns 

108 ------- 

109 basisList : `list` of `lsst.afw.math.kernel.FixedKernel` 

110 List of basis kernels. For each degree value ``n`` in ``config.basisDegGauss`` (n+2)(n+1)/2 kernels 

111 are generated and appended to the list in the order of the polynomial parameter number. 

112 See `lsst.afw.math.polynomialFunction2D` documentation for more details. 

113 

114 Raises 

115 ------ 

116 RuntimeError 

117 - if ``config.kernelBasisSet`` is not equal to "alard-lupton" 

118 ValueError 

119 - if ``config.kernelSize`` is even 

120 - if the number of Gaussians and the number of given 

121 sigma values are not equal or 

122 - if the number of Gaussians and the number of given 

123 polynomial degree values are not equal 

124 

125 Notes 

126 ----- 

127 The polynomial functions (``f``) are always evaluated in the -1.0, +1.0 range in both x, y directions, 

128 edge to edge, with ``f(0,0)`` evaluated at the kernel center pixel, ``f(-1.0,-1.0)`` at the kernel 

129 ``(0,0)`` pixel. They are not scaled by the sigmas of the Gaussians. 

130 

131 Base Gaussian widths (sigmas in pixels) of the kernels are determined as: 

132 - If not all fwhm parameters are provided or ``config.scaleByFwhm==False`` 

133 then ``basisSigmaGauss`` is used. If ``basisSigmaGauss`` is not 

134 provided, then ``config.alardSigGauss`` is used. In both cases, the 

135 length of sigmas must be equal to ``config.alardNGauss``. 

136 - If ``targetFwhmPix<referenceFwhmPix`` (normal convolution): 

137 First sigma ``Sig_K`` is determined to satisfy: ``Sig_reference**2 = Sig_target**2 + Sig_K**2``. 

138 If it's larger than ``config.alardMinSig * config.alardGaussBeta``, make it the 

139 second kernel. Else make it the smallest kernel, unless only 1 kernel is asked for. 

140 - If ``referenceFwhmPix < targetFwhmPix`` (deconvolution): 

141 Define the progression of Gaussians using a 

142 method to derive a deconvolution sum-of-Gaussians from it's 

143 convolution counterpart. [1]_ Only use 3 since the algorithm 

144 assumes 3 components. 

145 

146 **Metadata fields** 

147 

148 ALBasisNGauss : `int` 

149 The number of base Gaussians in the AL basis functions. 

150 ALBasisDegGauss : `list` of `int` 

151 Polynomial order of spatial modification of the base Gaussian functions. 

152 ALBasisSigGauss : `list` of `float` 

153 Sigmas in pixels of the base Gaussians. 

154 ALKernelSize : `int` 

155 Kernel stamp size is (ALKernelSize pix, ALKernelSize pix). 

156 ALBasisMode : `str`, either of ``config``, ``convolution``, ``deconvolution`` 

157 Indicates whether the config file values, the convolution or deconvolution algorithm 

158 was used to determine the base Gaussian sigmas and the kernel stamp size. 

159 

160 References 

161 ---------- 

162 .. [1] Ulmer, W.: Inverse problem of linear combinations of Gaussian convolution kernels 

163 (deconvolution) and some applications to proton/photon dosimetry and image 

164 processing. http://iopscience.iop.org/0266-5611/26/8/085002 Equation 40 

165 """ 

166 

167 if config.kernelBasisSet != "alard-lupton": 

168 raise RuntimeError("Cannot generate %s basis within generateAlardLuptonBasisList" % 

169 config.kernelBasisSet) 

170 

171 kernelSize = config.kernelSize 

172 fwhmScaling = config.kernelSizeFwhmScaling 

173 basisNGauss = config.alardNGauss 

174 basisGaussBeta = config.alardGaussBeta 

175 basisMinSigma = config.alardMinSig 

176 if basisDegGauss is None: 

177 basisDegGauss = config.alardDegGauss 

178 if basisSigmaGauss is None: 

179 basisSigmaGauss = config.alardSigGauss 

180 

181 if len(basisDegGauss) != basisNGauss: 

182 raise ValueError("len(basisDegGauss) != basisNGauss : %d vs %d" % (len(basisDegGauss), basisNGauss)) 

183 if len(basisSigmaGauss) != basisNGauss: 

184 raise ValueError("len(basisSigmaGauss) != basisNGauss : %d vs %d" % 

185 (len(basisSigmaGauss), basisNGauss)) 

186 if (kernelSize % 2) != 1: 

187 raise ValueError("Only odd-sized Alard-Lupton bases allowed") 

188 

189 logger = Log.getLogger("ip.diffim.generateAlardLuptonBasisList") 

190 if (targetFwhmPix is None) or (referenceFwhmPix is None) or (not config.scaleByFwhm): 

191 logger.info("PSF sigmas are not available or scaling by fwhm disabled, " 

192 "falling back to config values") 

193 if metadata is not None: 

194 metadata.add("ALBasisNGauss", basisNGauss) 

195 metadata.add("ALBasisDegGauss", basisDegGauss) 

196 metadata.add("ALBasisSigGauss", basisSigmaGauss) 

197 metadata.add("ALKernelSize", kernelSize) 

198 metadata.add("ALBasisMode", "config") 

199 

200 return diffimLib.makeAlardLuptonBasisList(kernelSize//2, basisNGauss, basisSigmaGauss, basisDegGauss) 

201 

202 targetSigma = targetFwhmPix / sigma2fwhm 

203 referenceSigma = referenceFwhmPix / sigma2fwhm 

204 logger.debug("Generating matching bases for sigma %.2f pix -> %.2f pix", targetSigma, referenceSigma) 

205 

206 # Modify the size of Alard Lupton kernels based upon the images FWHM 

207 # 

208 # Note the operation is : template.x.kernel = science 

209 # 

210 # Assuming the template and science image Psfs are Gaussians with 

211 # the Fwhm above, Fwhm_T **2 + Fwhm_K **2 = Fwhm_S **2 

212 # 

213 if targetSigma == referenceSigma: 

214 # Leave defaults as-is 

215 logger.debug("Target and reference psf fwhms are equal, falling back to config values") 

216 basisMode = "config" 

217 elif referenceSigma > targetSigma: 

218 # Normal convolution 

219 

220 # First Gaussian has the sigma that comes from the convolution 

221 # of two Gaussians : Sig_S**2 = Sig_T**2 + Sig_K**2 

222 # 

223 # If it's larger than basisMinSigma * basisGaussBeta, make it the 

224 # second kernel. Else make it the smallest kernel. Unless 

225 # only 1 kernel is asked for. 

226 logger.debug("Reference psf fwhm is the greater, normal convolution mode") 

227 basisMode = "convolution" 

228 kernelSigma = np.sqrt(referenceSigma**2 - targetSigma**2) 

229 if kernelSigma < basisMinSigma: 

230 kernelSigma = basisMinSigma 

231 

232 basisSigmaGauss = [] 

233 if basisNGauss == 1: 

234 basisSigmaGauss.append(kernelSigma) 

235 nAppended = 1 

236 else: 

237 if (kernelSigma/basisGaussBeta) > basisMinSigma: 

238 basisSigmaGauss.append(kernelSigma/basisGaussBeta) 

239 basisSigmaGauss.append(kernelSigma) 

240 nAppended = 2 

241 else: 

242 basisSigmaGauss.append(kernelSigma) 

243 nAppended = 1 

244 

245 # Any other Gaussians above basisNGauss=1 come from a scaling 

246 # relationship: Sig_i+1 / Sig_i = basisGaussBeta 

247 for i in range(nAppended, basisNGauss): 

248 basisSigmaGauss.append(basisSigmaGauss[-1]*basisGaussBeta) 

249 

250 kernelSize = int(fwhmScaling * basisSigmaGauss[-1]) 

251 kernelSize += 0 if kernelSize%2 else 1 # Make sure it's odd 

252 kernelSize = min(config.kernelSizeMax, max(kernelSize, config.kernelSizeMin)) 

253 

254 else: 

255 # Deconvolution; Define the progression of Gaussians using a 

256 # method to derive a deconvolution sum-of-Gaussians from it's 

257 # convolution counterpart. Only use 3 since the algorithm 

258 # assumes 3 components. 

259 # 

260 # http://iopscience.iop.org/0266-5611/26/8/085002 Equation 40 

261 

262 # Use specializations for deconvolution 

263 logger.debug("Target psf fwhm is the greater, deconvolution mode") 

264 basisMode = "deconvolution" 

265 basisNGauss = config.alardNGaussDeconv 

266 basisMinSigma = config.alardMinSigDeconv 

267 

268 kernelSigma = np.sqrt(targetSigma**2 - referenceSigma**2) 

269 if kernelSigma < basisMinSigma: 

270 kernelSigma = basisMinSigma 

271 

272 basisSigmaGauss = [] 

273 if (kernelSigma/basisGaussBeta) > basisMinSigma: 

274 basisSigmaGauss.append(kernelSigma/basisGaussBeta) 

275 basisSigmaGauss.append(kernelSigma) 

276 nAppended = 2 

277 else: 

278 basisSigmaGauss.append(kernelSigma) 

279 nAppended = 1 

280 

281 for i in range(nAppended, basisNGauss): 

282 basisSigmaGauss.append(basisSigmaGauss[-1]*basisGaussBeta) 

283 

284 kernelSize = int(fwhmScaling * basisSigmaGauss[-1]) 

285 kernelSize += 0 if kernelSize%2 else 1 # Make sure it's odd 

286 kernelSize = min(config.kernelSizeMax, max(kernelSize, config.kernelSizeMin)) 

287 

288 # Now build a deconvolution set from these sigmas 

289 sig0 = basisSigmaGauss[0] 

290 sig1 = basisSigmaGauss[1] 

291 sig2 = basisSigmaGauss[2] 

292 basisSigmaGauss = [] 

293 for n in range(1, 3): 

294 for j in range(n): 

295 sigma2jn = (n - j)*sig1**2 

296 sigma2jn += j * sig2**2 

297 sigma2jn -= (n + 1)*sig0**2 

298 sigmajn = np.sqrt(sigma2jn) 

299 basisSigmaGauss.append(sigmajn) 

300 

301 basisSigmaGauss.sort() 

302 basisNGauss = len(basisSigmaGauss) 

303 basisDegGauss = [config.alardDegGaussDeconv for x in basisSigmaGauss] 

304 

305 if metadata is not None: 

306 metadata.add("ALBasisNGauss", basisNGauss) 

307 metadata.add("ALBasisDegGauss", basisDegGauss) 

308 metadata.add("ALBasisSigGauss", basisSigmaGauss) 

309 metadata.add("ALKernelSize", kernelSize) 

310 metadata.add("ALBasisMode", basisMode) 

311 

312 logger.debug("basisSigmaGauss: %s basisDegGauss: %s", 

313 ','.join(['{:.1f}'.format(v) for v in basisSigmaGauss]), 

314 ','.join(['{:d}'.format(v) for v in basisDegGauss])) 

315 

316 return diffimLib.makeAlardLuptonBasisList(kernelSize//2, basisNGauss, basisSigmaGauss, basisDegGauss)