Introduction to MOMFBD a short overview Mats Löfdahl Institute for Solar Physics Stockholm University 1st CASSDA-SOLARNET Workshop Freiburg 18-20 February, 2014 Mats Löfdahl (Institute for Solar Physics) Introduction to MOMFBD Freiburg 2014-02-18–20 1 / 13 Multi-Object Multi-Frame Blind Deconvolution 1 What is MOMFBD? MOMFBD – Deconvolution MOMFBD – Blind Deconvolution MOMFBD – Multiple Frames MOMFBD – Phase Diversity MOMFBD – Multiple Objects 2 History 3 Problems and solutions Alignment Truncated wavefront expansion Model mismatch Mats Löfdahl (Institute for Solar Physics) Introduction to MOMFBD Freiburg 2014-02-18–20 2 / 13 Multi-Object Multi-Frame Blind Deconvolution 1 What is MOMFBD? MOMFBD – Deconvolution MOMFBD – Blind Deconvolution MOMFBD – Multiple Frames MOMFBD – Phase Diversity MOMFBD – Multiple Objects 2 History 3 Problems and solutions Alignment Truncated wavefront expansion Model mismatch Mats Löfdahl (Institute for Solar Physics) Introduction to MOMFBD Freiburg 2014-02-18–20 2 / 13 Multi-Object Multi-Frame Blind Deconvolution 1 What is MOMFBD? MOMFBD – Deconvolution MOMFBD – Blind Deconvolution MOMFBD – Multiple Frames MOMFBD – Phase Diversity MOMFBD – Multiple Objects 2 History 3 Problems and solutions Alignment Truncated wavefront expansion Model mismatch Mats Löfdahl (Institute for Solar Physics) Introduction to MOMFBD Freiburg 2014-02-18–20 2 / 13 What is MOMFBD? MOMFBD – Deconvolution Blurring and unblurring Convolution d =f ∗s Mats Löfdahl (Institute for Solar Physics) = Introduction to MOMFBD ∗ Freiburg 2014-02-18–20 3 / 13 What is MOMFBD? MOMFBD – Deconvolution Blurring and unblurring Convolution d =f ∗s = ∗ = ∗−1 Deconvolution f = d ∗−1 s Mats Löfdahl (Institute for Solar Physics) Introduction to MOMFBD Freiburg 2014-02-18–20 3 / 13 What is MOMFBD? MOMFBD – Blind Deconvolution BD: Joint estimation of object and aberrations Image: Image = mix of two unknown quantities Mats Löfdahl (Institute for Solar Physics) Introduction to MOMFBD How do they separate? Freiburg 2014-02-18–20 4 / 13 What is MOMFBD? MOMFBD – Blind Deconvolution BD: Joint estimation of object and aberrations Object: PSF: Perfect optics and fuzzy object? Mats Löfdahl (Institute for Solar Physics) Introduction to MOMFBD Or... Freiburg 2014-02-18–20 4 / 13 What is MOMFBD? MOMFBD – Blind Deconvolution BD: Joint estimation of object and aberrations Object: PSF: Stellar object and weird PSF? Mats Löfdahl (Institute for Solar Physics) Introduction to MOMFBD Or... Freiburg 2014-02-18–20 4 / 13 What is MOMFBD? MOMFBD – Blind Deconvolution BD: Joint estimation of object and aberrations Object: PSF: Constrain the solution and find ... A star and a proper PSF! Mats Löfdahl (Institute for Solar Physics) Introduction to MOMFBD Freiburg 2014-02-18–20 4 / 13 What is MOMFBD? MOMFBD – Multiple Frames Data collection model: Multiple samples of seeing Extended target Turbulence ☛ ❘ Optics ❄ Filter ✗ Collected images: Detector ... t = 1, 2, Mats Löfdahl (Institute for Solar Physics) 3, ... T – Varying seeing, one object Introduction to MOMFBD Freiburg 2014-02-18–20 5 / 13 What is MOMFBD? MOMFBD – Multiple Frames Maths Image formation model dj = f ∗ sj + nj sj = | F −1 {Pj Image }|2 PSF Pj = A · exp(iφj ) P ˆ φj ≈ M m αmj ψm = φj Pupil function Pupil phase Image = Mats Löfdahl (Institute for Solar Physics) ∗ + Introduction to MOMFBD Freiburg 2014-02-18–20 6 / 13 What is MOMFBD? MOMFBD – Multiple Frames Maths Image formation model dj = f ∗ sj + nj sj = | F −1 {Pj }|2 Pj = A · exp(iφj ) P ˆ φj ≈ M m αmj ψm = φj Image PSF Pupil function Pupil phase PSF and pupil function −1 = F · exp i · Mats Löfdahl (Institute for Solar Physics) 2 Introduction to MOMFBD Freiburg 2014-02-18–20 6 / 13 What is MOMFBD? MOMFBD – Multiple Frames Maths Image formation model dj = f ∗ sj + nj sj = | F −1 {Pj Image }|2 PSF Pj = A · exp(iφj ) P ˆ φj ≈ M m αmj ψm = φj Pupil function Pupil phase Pupil phase ≈ α4 · + α5 · Mats Löfdahl (Institute for Solar Physics) + α6 · Introduction to MOMFBD + · · · + αM · = Freiburg 2014-02-18–20 6 / 13 What is MOMFBD? MOMFBD – Multiple Frames Maths Model fitting Image formation model dj = f ∗ sj + nj sj = | F −1 {Pj Image }|2 PSF Pj = A · exp(iφj ) P ˆ φj ≈ M m αmj ψm = φj Pupil function Pupil phase Minimize difference between data and model data: P P minα |dj − ˆf ∗ sˆj |2 j pixels Pupil phase ≈ α4 · + α5 · + α6 · + · · · + αM · = Non-linear optimization methods, linear equality constraints Mats Löfdahl (Institute for Solar Physics) Introduction to MOMFBD Freiburg 2014-02-18–20 6 / 13 What is MOMFBD? MOMFBD – Multiple Frames Maths Model fitting Image formation model dj = f ∗ sj + nj sj = | F −1 {Pj Image }|2 PSF Pj = A · exp(iφj ) P ˆ φj ≈ M m αmj ψm = φj Pupil function Pupil phase Minimize difference between data and model data: P P minα |dj − ˆf ∗ sˆj |2 j pixels Pupil phase ≈ α4 · + α5 · + α6 · + · · · + αM · = Karhunen–Loève functions statistically uncorrelated for atmosphere Mats Löfdahl (Institute for Solar Physics) Introduction to MOMFBD Freiburg 2014-02-18–20 6 / 13 What is MOMFBD? MOMFBD – Phase Diversity Data collection model: Multiple samples of seeing Extended target Turbulence ☛ ❘ Optics ❄ Filter ✗ Collected images: Detector ... t = 1, 2, Mats Löfdahl (Institute for Solar Physics) 3, ... T – Varying seeing, one object Introduction to MOMFBD Freiburg 2014-02-18–20 7 / 13 What is MOMFBD? MOMFBD – Phase Diversity Data collection model: Multiple samples of seeing with PD Extended target Turbulence ☛ ❘ Shutter Optics ❄ Beam splitter ❯ Filter ☛ ✗ Collected images: d = 0: ... d = 1: ... t = 1, 2, Mats Löfdahl (Institute for Solar Physics) 3, ... ✐ Detector Two cameras w/ identical seeing, intentional focus difference T – Varying seeing, one object Introduction to MOMFBD Freiburg 2014-02-18–20 7 / 13 MOMFBD – Multiple Objects What is MOMFBD? Data collection model: Multiple samples of seeing with MO Extended target Shutter Turbulence ☛ ❘ Optics ❄ Beam splitter ❯ ❯ Filter i = 1 Filter i = 2 Collected images: i = 1: ... i = 2: ... t = 1, 2, Mats Löfdahl (Institute for Solar Physics) 3, ... ✗ ❨ Detector Two cameras w/ identical seeing T – Varying seeing, two objects Introduction to MOMFBD Freiburg 2014-02-18–20 8 / 13 History MOMFBD time line MFBD and MOMFBD Phase Diversity R.A. Gonsalves. Optical Engineering, 21(5):829–832, 1982. T.J. Schulz. JOSA A, 10:1064–1073, 1993. R.G. Paxman, T.J. Schulz, and J.R. Fienup. JOSA A, 9(7):1072–1085, 1992a. M.G. Löfdahl. Proc. SPIE, 4792:146, 2002. R.G. Paxman, T.J. Schulz, and J.R. Fienup. OSA Technical Digest Series, 11:5, 1992b. M.G. Löfdahl and G.B. Scharmer. A&A Suppl. Ser., 107:243, 1994. R.G. Paxman, J.H. Seldin, M.G. Löfdahl, G.B. Scharmer, and C.U. Keller. ApJ, 466:1087, 1996. M. van Noort, L. Rouppe van der Voort, and M.G. Löfdahl. Solar Physics, 228(1–2):191, 2005. M.J. van Noort and L.H.M. Rouppe van der Voort. ApJ, 489:429, 2008. R. Schnerr, J. de la Cruz Rodríguez, and M. van Noort. A&A, 534:A45, 2011. V.M.J. Henriques. A&A, 548:A114, 2012. Phase Diversity Mats Löfdahl (Institute for Solar Physics) Introduction to MOMFBD Freiburg 2014-02-18–20 9 / 13 History MOMFBD time line MFBD and MOMFBD Phase Diversity R.A. Gonsalves. Optical Engineering, 21(5):829–832, 1982. T.J. Schulz. JOSA A, 10:1064–1073, 1993. R.G. Paxman, T.J. Schulz, and J.R. Fienup. JOSA A, 9(7):1072–1085, 1992a. M.G. Löfdahl. Proc. SPIE, 4792:146, 2002. R.G. Paxman, T.J. Schulz, and J.R. Fienup. OSA Technical Digest Series, 11:5, 1992b. M.G. Löfdahl and G.B. Scharmer. A&A Suppl. Ser., 107:243, 1994. R.G. Paxman, J.H. Seldin, M.G. Löfdahl, G.B. Scharmer, and C.U. Keller. ApJ, 466:1087, 1996. M. van Noort, L. Rouppe van der Voort, and M.G. Löfdahl. Solar Physics, 228(1–2):191, 2005. M.J. van Noort and L.H.M. Rouppe van der Voort. ApJ, 489:429, 2008. R. Schnerr, J. de la Cruz Rodríguez, and M. van Noort. A&A, 534:A45, 2011. V.M.J. Henriques. A&A, 548:A114, 2012. Phase Diversity, Solar Phase Diversity Mats Löfdahl (Institute for Solar Physics) Introduction to MOMFBD Freiburg 2014-02-18–20 9 / 13 History MOMFBD time line MFBD and MOMFBD Phase Diversity R.A. Gonsalves. Optical Engineering, 21(5):829–832, 1982. T.J. Schulz. JOSA A, 10:1064–1073, 1993. R.G. Paxman, T.J. Schulz, and J.R. Fienup. JOSA A, 9(7):1072–1085, 1992a. M.G. Löfdahl. Proc. SPIE, 4792:146, 2002. R.G. Paxman, T.J. Schulz, and J.R. Fienup. OSA Technical Digest Series, 11:5, 1992b. M.G. Löfdahl and G.B. Scharmer. A&A Suppl. Ser., 107:243, 1994. R.G. Paxman, J.H. Seldin, M.G. Löfdahl, G.B. Scharmer, and C.U. Keller. ApJ, 466:1087, 1996. M. van Noort, L. Rouppe van der Voort, and M.G. Löfdahl. Solar Physics, 228(1–2):191, 2005. M.J. van Noort and L.H.M. Rouppe van der Voort. ApJ, 489:429, 2008. R. Schnerr, J. de la Cruz Rodríguez, and M. van Noort. A&A, 534:A45, 2011. V.M.J. Henriques. A&A, 548:A114, 2012. Phase Diversity, Solar Phase Diversity, MFBD & MOMFBD Mats Löfdahl (Institute for Solar Physics) Introduction to MOMFBD Freiburg 2014-02-18–20 9 / 13 History MOMFBD time line MFBD and MOMFBD Phase Diversity R.A. Gonsalves. Optical Engineering, 21(5):829–832, 1982. T.J. Schulz. JOSA A, 10:1064–1073, 1993. R.G. Paxman, T.J. Schulz, and J.R. Fienup. JOSA A, 9(7):1072–1085, 1992a. M.G. Löfdahl. Proc. SPIE, 4792:146, 2002. R.G. Paxman, T.J. Schulz, and J.R. Fienup. OSA Technical Digest Series, 11:5, 1992b. M.G. Löfdahl and G.B. Scharmer. A&A Suppl. Ser., 107:243, 1994. R.G. Paxman, J.H. Seldin, M.G. Löfdahl, G.B. Scharmer, and C.U. Keller. ApJ, 466:1087, 1996. M. van Noort, L. Rouppe van der Voort, and M.G. Löfdahl. Solar Physics, 228(1–2):191, 2005. M.J. van Noort and L.H.M. Rouppe van der Voort. ApJ, 489:429, 2008. R. Schnerr, J. de la Cruz Rodríguez, and M. van Noort. A&A, 534:A45, 2011. V.M.J. Henriques. A&A, 548:A114, 2012. Phase Diversity, Solar Phase Diversity, MFBD & MOMFBD, MOMFBD strategies Mats Löfdahl (Institute for Solar Physics) Introduction to MOMFBD Freiburg 2014-02-18–20 9 / 13 Problems and solutions Alignment Pinhole calibrated camera alignment arcseconds 10 5 a 0 b arcseconds 10 5 c 0 0 d 5 arcseconds 0 5 arcseconds a. LCP; b. RCP; c. MOMFBD magnetogram; d. Traditional magnetogram; note artifacts Mats Löfdahl (Institute for Solar Physics) Introduction to MOMFBD Freiburg 2014-02-18–20 10 / 13 Problems and solutions Single WB object Alignment D iv er sit y k van Noort et al. (2005) Fe 630.2 Wide Object i Fe 630.2 LCP Fe 630.2 RCP 1 Realization t 9 11 19 All WB frames ⇒ single restored image NB frames co-restored with some WB frames Different NB states aligned by the common WB image Mats Löfdahl (Institute for Solar Physics) Introduction to MOMFBD Freiburg 2014-02-18–20 11 / 13 Problems and solutions Single WB object Alignment Extra WB objects Henriques (2012) D iv er sit y k van Noort et al. (2005) Fe 630.2 Wide Object i Fe 630.2 LCP Fe 630.2 RCP 1 Realization t 9 11 19 All WB frames ⇒ single restored image NB frames co-restored with some WB frames Different NB states aligned by the common WB image Mats Löfdahl (Institute for Solar Physics) Introduction to MOMFBD Allow for residual misalignment Freiburg 2014-02-18–20 11 / 13 Encir 0.4 0.4 0.4 Problems and solutions 0.2 5 cm Truncated wavefront expansion 0.2 7 cm 0.2 5 cm A finite number of wavefront modes 0.0 0 1 2 3 0.0 5 10 15 r0 / 1 cm r / 1" Fig. 2: Encircled PSF energy for different r0 as indicated in the figure. Red: PSFs based on S ; Blue: PSFs based on Sˆ ; Black: diffraction limited PSF; Black dashed: 90% level. 20 25 Fig. 3: Strehl ratios as a function of r0 . Solid line: Eq. (11); Plus (+) symbols: PSFs based on S ; Cross (×) symbols: PSFs based on Sˆ . 0.0 0.0 0.2 0.4 0.6 0.8 1.0 u Fig. 4: Power spectra (angular averages) of S (red) and Sˆ (blue) for r0 as indicated. Perfect correction of 36 KL modes restores resolution, not contrast 13.29% 12.78% 11.83% 9.77% 7.23% 4.63% 14.51% 14.51% 14.50% 14.48% 14.41% 14.29% 14.52% = ∞ Far left: Original 25 cmimage. Top: Low-pass 20 cm images degraded 15 cmby high-order aberrations, 10 cm Fig. 5:r0Images. i.e., by S7, cm corresponding to5r0cm =25, 20, 15, 10, 7 and 5 cm, resp. (same layout as Fig. 1). Bottom: Degraded images compensated by use of the method described, i.e., by Sˆ . All images are scaled between min and max of the original image and low-pass filtered to 90% of the SST diffraction limit. The numbers above and below the image tiles are the RMS contrasts in percent of the mean intensity (100 × RMS/mean). Solution The synthetic images were correction processed with the based MOMFBD on parameters, without and withstatistics. added noise). In this section Post-restoration atmospheric Not in we program in various ways (MFBD or JPDS, different numbers of summarize the results of these simulations. !! pipeline yet. Scharmer & Löfdahl (2010) realizations, different subfield sizes (256 pixels = 15 , 128 pixels !! !! = 7. 6, 80 pixels = 4. 7), different number of estimated wavefront Mats Löfdahl (Institute for Solar Physics) Figure 6 shows RMS intensity errors of restored images using different techniques (MFBD, JPDS), number of aberration parameters (M=35, 50 and 100) without and with Sˆ compensa- Introduction to MOMFBD Freiburg 2014-02-18–20 12 / 13 Problems and solutions Model mismatch Model mismatch hurts the fit Pupils Non-common paths Telescope FL Field stop/pinholes CT DC DM Blue path TM RL Prefilter Shutter E LR Cs SL PB E HR Other problems AO WFS Optical parameters Not used PI al F Du Finite exposures WB R NB T NB ISP CR Flats: time-dependent, polarized, IR Symptoms Under-estimated wavefronts ⇒ under-corrected images Small-scale image artifacts amplified by deconvolution Mats Löfdahl (Institute for Solar Physics) Introduction to MOMFBD Freiburg 2014-02-18–20 13 / 13
© Copyright 2025 ExpyDoc