2. Observation: instruments & sensors

Transcription

1 Master ISTI / PARI / IV Introduction to Astronomical Image Processing 2. Observation: instruments & sensors André Jalobeanu LSIIT / MIV / PASEO group Dec lsiit-miv.u-strasbg.fr/paseo PASEO

2 Observation: instruments & sensors Telescopes (physics, examples) Problems using telescopes Diffraction limit Atmospheric turbulence and opacity Complex hardware solutions Interferometry (multiple instruments) Active and Adaptive optics 2D sensors Optical imaging (UV, visible, IR), Radio, X, Gamma Multispectral imaging with 2D sensors 3D sensors: integral field spectroscopy (IFS) Hyperspectral imaging: fibers, slicers, lensarrays Summary: the image acquisition chain

3 Astronomical telescopes: an overview Grasp image formation concepts (optical and non-optical devices) Understand the problems related to diffraction and atmosphere Be aware of advanced hardware solutions specific to astronomy

4 Optical telescopes: introduction (UV, visible, IR) 50cm refractor (Nice) 20cm amateur telescope 8.2m Very Large Telescope (VLT) #1 2.5m Nordic Optical Telescope 2.4m Hubble Space Telescope (HST)

5 Optical reflecting telescopes: principle Optical tube assembly Primary mirror: Light collector Parabolic shape (in general) Secondary mirror Reflection (get the light out) Correction in some cases Different designs, various applications Newton Cassegrain Maksutov Schmidt Camera

6 X-ray telescopes Grazing angle reflection only! Nested metallic mirrors Paraboloid+Hyperboloid surfaces Space telescopes X-rays are deflected by radiation belts Problems Focusing on the entire field of view High-energy particles in deep space 58 nested X-ray grazing incidence mirrors (XMM-Newton, ESA)

7 Radio Telescopes Parabolic antennas Metallic dishes, rough surface comp. to optical telescopes! Aperture synthesis Mix signals from a collection of instruments (or use Earth rotation) Each instrument provides an image sample in the Fourier space NRAO Very Large Array (VLA) Problem: Sparse sampling of the Fourier space

8 Diffraction and finite-size instruments Diffraction (monochromatic): Ideal instrument, circular aperture D Airy pattern point spread function (PSF) hat modulation transfer function (MTF) larger aperture size, sharper image... and higher cost! Airy pattern resolution: 1.22λ/D MTF PSF MTF = autocorrelation of the pupil function Pupil shape and MTF 10 cm aperture Damian Peach 30 cm aperture

9 Optical aberrations and distortions Monochromatic aberrations Spherical aberrations Astigmatism Coma 30 cm aperture top: no spherical aberration bottom: lambda/2 aberration Damian Peach Geometric distortions Barrel/cushion distortions Chromatic aberrations Dispersion through lenses (refractors)

10 Atmospheric opacity

11 Atmospheric extinction and refraction Extinction Airmass depends on the zenith angle Molecular absorption Rayleigh scattering Aerosol scattering Wavelength-dependent effect! Refraction Varying atmospheric refraction index and lens-shaped atmosphere Zenith angle-dependent (airmass) Wavelength-dependent (dispersion)

12 Sky background Atmosphere Sunlight scattering (dust, aerosols) Stray lights (light pollution) Sky emission lines Pinatubo effect Europe by night Interplanetary medium Sunlight scattering (zodiacal light) Zodiacal light

13 Atmospheric turbulence: short exposure Short exposures Distorted wave-front: phase error Kolmogorov turbulence statistics Speckles (Max Planck Institute) Binary star

14 Atmospheric turbulence - long exposure Long exposures: average over the (long) integration time! Long exposure modulation transfer function: MTF s (ν) = e 3.44 (λ f ν/r 0) 5/3 [1 b (λ f ν/d) 1/3 ] ν spatial frequency λ wavelength f focal length D aperture diameter br constant (1 for far-field) r0 Fried's seeing diameter Equivalent diameter: seeing diameter r0 Long exposure MTF vs. telescope MTF

15 Aperture Synthesis Interferometry hardware solution to the diffraction limit Go beyond the diffraction limit? Build larger telescopes... or use multiple telescopes simultaneously! Interference (2 telescopes): Produce interference fringes (contrast+phase) In practice: measure the amplitude, phase from phase closure... (random phase fluctuations due to atmospheric turbulence) Sample the Fourier (u,v) space

16 Interferometry - the VLTI Angular resolution: longest baseline Problems: Sparse sampling of the Fourier space Noise (amplitude) Phase determination

17 Active and Adaptive Optics hardware solutions to atmospheric turbulence Wave-front sensor: measure the phase error Deformable mirror: correct the wave-front

18 Sensors: image sampling in 2D/3D Understand how 2D sensors provide single band or multispectral images Have an idea of how 3D sensors work (integral field spectroscopy) Understand the problems related to these sensors (e.g. noise)

19 2D sensors Optical imaging (UV, visible, IR) Optical detectors (from UV to IR): CCD vs. CMOS: image transfer image on the focal plane Problems: Quantum Efficiency (QE) Poor in some wavelength ranges, wavelength-dep: Noise (stochastic errors) Photon counting, readout, cosmic... Bias (systematic errors) Dark current, non-linearity, bad pixels... Blur caused by charge diffusion CCD matrix and pixel integration Sensitive area (pixel)

20 Noise in electronic detectors stochastic errors Noise: stochastic process Observation: realization of a random variable Assumptions : Several additive processes, zero mean Stochastic independence between pixels (white noise) Stationary process (although parameters may be non-stationary) Probabilistic model: P(d m) conditional pdf value of the observed pixel mean Counting: Poisson Readout: Gaussian Background (thermal, sky): ~ Poisson (shift m), cosmic rays: impulse noise Quantization: uniform (bounded) Simplification: high photon count Gaussian noise (variance = m d)

21 Bias in electronic detectors systematic errors Dark current Background current in the detector, pixel-dependent (both current and QE decrease with temperature) Non-linearity Non-linearity of the amplification process Pixel saturation, blooming Pixel-dependent sensitivity QE depends on each pixel: flat-field image Bad pixels Hot pixels: close to saturation Cold pixels: low efficiency or various electronic problems Cross-talk Charge diffusion between pixels (usually during transfer)

22 X-ray sensors Photon imaging Record time, position, energy Use special CCD detectors (PN-CCD) Over 90% QE, range keV Problems: Photon-counting noise Particle noise (tracks) Geminga (neutron star) X-ray CCD for XMM (MPE)

23 Gamma sensors Use the Compton effect 2 layers: scintillator, photon detector Record time, location, energy Orientation from location on each layer CGRO (NASA) Problems: Inaccurate orientation determination Photon-counting noise

24 Radio sensors Radio antenna technology (EM wave current) Problems: Sparse sampling of the Fourier space Correlated noise on reconstructed images! Galaxy 3C353, VLA 3.6cm (NRAO)

25 Multiband imaging with 2D sensors Filter the light between telescope and sensor Classical absorption filters (bandpass) Fabry-Pérot interferometer (narrow band) tunable filters: var. wavelength and bandwidth Problems: Non-simultaneous observations: image registration needed! Data cube scrambling: interference rings have to be removed Time consuming for large number of bands (usually <30)

26 3D sensors: Integral Field Spectroscopy (IFS) one spectrum for each pixel ( spaxel ) Fiber-fed spectrograph Optical fibers in the focal plane: rearrange spatial information (field to slit mapping) Feed a dispersion system and record spectra on a CCD sensor Lenslet arrays: enhance sampling and sensitivity GMOS Integral Field Unit (Durham University)

27 IFS - lenslet arrays Array of small lenses Solve problems related to sampling/sensitivity (gaps btw fibers) No fiber transmission problem Drawbacks: lower spectral resolution, scrambling on the sensor OASIS optical layout (CFHT)

28 IFS - Image slicers, summary Slicer mirror system-fed spectrograph Cut focal plane into multiple slices Solve problems related to fiber transmission Problems: (IFS in general) Few spaxels (spatial elements): max 80x80 (300x300 in ) Unscrambling (CCD to field): correlated noise from resampling Dimensionality & redundancy IFS Summary GNIRS Integral Field Unit (Durham University)

29 Optical image acquisition chain schematics deep-sky object values support continuous discrete deterministic turbulence, opacity atmosphere stochastic optics aberrations diffraction telescope counting, thermal, cosmics sensor motion, defects detector physics geometry, physics scan amp. readout noise conversion A/D Observed Image

30 A simplified optical observation model deterministic values stochastic values noise T + Natural scene Blur Sampling Trans. observed image continuous support discrete support Convolution by PSF (Point Spread Function) multiply by Dirac array Pixel value transform additive noise probabilistic model Simplification : Shift invariance mult. by MTF in freq. space Simplification : ~ Nyquist Simplification : Identity Simplification : Gaussian, white, stationary OTF(ξ,η) = MTF(ξ,η) PTF(ξ,η) F[Y] = MTF.F[X] + N variance σ 2

31 Modeling the blur origins and expressions Atmosphere : turbulence, diffusion Optics : diffraction, aberrations (defocus, spherical aberration,...) Sensor (geometry) : integration, motion (drift,...) Sensor (physics) : charge diffusion MTF 0 = MTF atm. MTF opt. MTF geom. MTF phys Global MTF Turbulence : [Kolmogorov 41] Diffusion: Truncated Gaussian Incoherent light optics - translation: sinc - oscillations Charge diffusion: Gaussian

32 Observation: instruments & sensors Further reading (web links) Various Wikipedia articles Adaptive optics tutorial at CTIO (turbulence & AO) VLTI General Description and tutorials ESO Instrumentation Hubble Nuts & Bolts Modeling the MTF and Noise Characteristics of Complex Image Formation Systems, Brian M. Bleeze (PhD Thesis): Modèles, estimation bayésienne et algorithmes pour la déconvolution d'images satellitaires et aériennes, Chap. 2 & References, André Jalobeanu (PhD Thesis, French)