Multi-band Deblending with scarlet Fred Moolekamp Princeton - PowerPoint PPT Presentation


 
 Multi-band Deblending with scarlet 
 Fred Moolekamp 
 Princeton University 
 LSST 2018 Blending Workshop 
 August 14, 2018 #lsst2018 #lsst2018 LSST Project and Community Workshop 2018 • Tucson • August 13 - 17 LSST Project and Community Workshop 2018 • Tucson • August 13 - 17 1

The problem - Blended scene: Image from HSC COSMOS dataset (courtesy Peter Melchior) #lsst2018 LSST Project and Community Workshop 2018 • Tucson • August 13 - 17 2

The problem - Use colors! Image from HSC COSMOS dataset (courtesy Peter Melchior) #lsst2018 LSST Project and Community Workshop 2018 • Tucson • August 13 - 17 3

scarlet : Basic Model Data Model Peak 2 Peak 0 Peak 1 Peak 3 Peak 4 #lsst2018 LSST Project and Community Workshop 2018 • Tucson • August 13 - 17 4

scarlet : Basic Model components pixels K ∑ Model for a blend: M = A k × S k A S bands k =1 model We want to minimize: K f ( A , S ) = 1 ( g A i ( S i ) ) ∑ 2 || Y − AS || 2 i ( A i ) + g S 2 + i =1 non-smooth constraints Data Use gradient decent and proximal operators to solve #lsst2018 LSST Project and Community Workshop 2018 • Tucson • August 13 - 17 5

Initialization - Build a symmetric, monotonic template at each peak - Use the monotonic templates to calculate the average SED Model at different iterations 0 1 5 27 #lsst2018 10 LSST Project and Community Workshop 2018 • Tucson • August 13 - 17 6

Constraints - SED Constraints - Non-negative - Sum to unity #lsst2018 LSST Project and Community Workshop 2018 • Tucson • August 13 - 17 7

Constraints Peak Ref - Morphology Constraints: - Non-negative - L0 sparsity Pixel - Soft symmetry: 2 ( S k , n − 1 + S † k , n − 1 ) S k , n = (1 − σ ) S k , n − 1 + σ - Monotonicity: - Each pixel must be less than or equal to the sum of its reference pixels - Only approximate constraint #lsst2018 LSST Project and Community Workshop 2018 • Tucson • August 13 - 17 8

180 degree Symmetry - separates most objects ? #lsst2018 LSST Project and Community Workshop 2018 • Tucson • August 13 - 17 9

Monotonicity - Forces flux to be non- increasing radially from the peak - Implemented as a projection operator ? onto the space of decreasing radial flux #lsst2018 LSST Project and Community Workshop 2018 • Tucson • August 13 - 17 10 10

PSF Deconvolution and Re-centering - PSF convolution and fractional pixel shifts use linear approximations - Makes the model more complicated by solvable k = source # i = pixel # b = band M b k = A b k T ki P b ii S ik Data Fully De-Convolved Model Model #lsst2018 LSST Project and Community Workshop 2018 • Tucson • August 13 - 17 11

Problems with Full PSF De-Convolution deblended galaxy - The constraints act on the deconvolved space (must be well sampled) - Full deconvolution = morphology under sampled - Results in artifacts in the models, for example: Artifacts due to blending Truth No PSF Full PSF Partial PSF #lsst2018 LSST Project and Community Workshop 2018 • Tucson • August 13 - 17 12

PSF Matching - Target PSF is a moffat - Create a difference kernel in each babd to match the moffat - The morphology in each band is convolved with the difference kernel #lsst2018 LSST Project and Community Workshop 2018 • Tucson • August 13 - 17 13

Conclusion - The last stack can now run scarlet using the MultibandDeblendTask and DeblendCoaddSourcesTask - scarlet can also be run as a stand alone package - See the documentation at http://scarlet.readthedocs.io - Tutorials will be shown at the “Hands on Deblending” workshop this afternoon - See the scarlet paper at https://arxiv.org/abs/1802.10157 - download scarlet at https://github.com/fred3m/scarlet #lsst2018 LSST Project and Community Workshop 2018 • Tucson • August 13 - 17 14