Image Deconvolution of Image Deconvolution of XMM-Newton DataXMM-Newton Data
Tao Song, Steve SembayTao Song, Steve Sembay
Dept. Physics & AstronomyDept. Physics & AstronomyUniversity of LeicesterUniversity of Leicester
OverviewOverview
• IntroductionIntroduction
• Richardson-Lucy AlgorithmRichardson-Lucy Algorithm
• IDL programIDL program
• ExamplesExamples
• Future worksFuture works
Vela PWNeChandra
EPIC-MOSDeconvolved
IntroductionIntroduction
• Observed images are usually degraded, i.e. the Observed images are usually degraded, i.e. the shape of a target will be distorted by the PSF.shape of a target will be distorted by the PSF.
• Image deconvolution is to recover the original Image deconvolution is to recover the original scene from the observed degraded data.scene from the observed degraded data.
• Two types of algorithms: empirical (e.g. CLEAN) Two types of algorithms: empirical (e.g. CLEAN) and theoretical (e.g. Richardson-Lucy)and theoretical (e.g. Richardson-Lucy)
• IDL software was developed to do image IDL software was developed to do image deconvolution on XMM-Newton datadeconvolution on XMM-Newton data
• Modified Richardson-Lucy algorithms and blind Modified Richardson-Lucy algorithms and blind deconvolution algorithm were testeddeconvolution algorithm were tested
Richardson-Lucy Richardson-Lucy DeconvolutionDeconvolution
)1( dxξ|xPxx
ΨξΨ rr1r
where dξξ|xPξΨx rrr
)2(' dxξ|xΨxx
PξP rr1r
)3(
1
1
2
2
n
iri
rii
n
where n is the number of pixels in a image.
IDL program – IDL program – main panelmain panel
IDL program – IDL program – detail paneldetail panel
ExamplesExamples – – P0401240501M1S001MIEVLI0000.FITP0401240501M1S001MIEVLI0000.FIT
PSF Data: PSF Data: PSF_M1_0a_VelaPSR_110208_i3_s1.fitsPSF_M1_0a_VelaPSR_110208_i3_s1.fits
•1.1’’ per pixel1.1’’ per pixel•FLAG == 0FLAG == 0•CCDNR == 1CCDNR == 1•PATTERN == 0PATTERN == 0
•50 Iterations50 Iterations22 = 1.80 = 1.80
ExamplesExamples – – P0401240501M1S001MIEVLI0000.FITP0401240501M1S001MIEVLI0000.FIT
Peak: 52277 (Y=244)FWHM: 3.14
(242.37~245.51)
Peak: 51398 (X=268)FWHM: 3.29
(266.70~269.99)
Peak: 14647 (Y=244)FWHM: 7.45
(240.28~247.73)
Peak: 14191 (X=269)FWHM: 8.32
(264.58~272.91)
ExamplesExamples – – P011108001M1S001MIEVLI000.FITP011108001M1S001MIEVLI000.FIT•1.1’’ per pixel1.1’’ per pixel•FLAG == 0FLAG == 0•CCDNR == 1CCDNR == 1•PATTERN == 0PATTERN == 0
•50 Iterations50 Iterations22 = 1.33 = 1.33
PSF Data: PSF Data: PSF_M1_0a_VelaPSR_110208_i3_s1.fitsPSF_M1_0a_VelaPSR_110208_i3_s1.fits
ExamplesExamples – – P011108001M1S001MIEVLI000.FITP011108001M1S001MIEVLI000.FIT
Peak: 35000 (Y=235)FWHM: 4.19
(232.96~237.15)
Peak: 30031 (X=257)FWHM: 4.39
(254.35~258.74)
Peak: 13385 (Y=235)FWHM: 14.18
(227.38~241.56)
Peak: 11595 (X=257)FWHM: 13.79
(249.44~263.23)
ExamplesExamples – blind deconvolution – blind deconvolution
22 = 1.20 = 1.20vs.vs.
22 = 1.33 = 1.33
22 = 1.24 = 1.24vs.vs.
22 = 1.80 = 1.80
ExamplesExamples – blind deconvolution – blind deconvolutionOutput strongly depends on the initial inputs, i.e. observed imageOutput strongly depends on the initial inputs, i.e. observed image and PSF and PSF
ExamplesExamples – blind deconvolution – blind deconvolution
A hint ?A hint ?(which one is (which one is more proper ?)more proper ?)
ExamplesExamples – blind deconvolution – blind deconvolution
22 = 1.77 vs. = 1.77 vs. 22 = 1.80 = 1.80
Peak: 56119 (Y=244)FWHM: 3.07
(242.38~245.45)
Peak: 54883 (X=268)FWHM: 3.19
(266.71~269.90)
Peak: 52277 (Y=244)FWHM: 3.14
(242.37~245.51)
Peak: 51398 (X=268)FWHM: 3.29
(266.70~269.99)
ExamplesExamples – blind deconvolution – blind deconvolution
22 = 1.33 vs. = 1.33 vs. 22 = 1.33 = 1.33
Peak: 36227 (Y=235)FWHM: 4.17
(232.94~237.10)
Peak: 31070 (X=257)FWHM: 4.38
(254.32~258.70)
Peak: 35000 (Y=235)FWHM: 4.19
(232.96~237.15)
Peak: 30031 (X=257)FWHM: 4.39
(254.35~258.74)
Future worksFuture works
• Apply image deconvolution on more Apply image deconvolution on more XMM-Newton DataXMM-Newton Data
• More tests on different PSFs (based on More tests on different PSFs (based on the outputs of blind deconvolution)the outputs of blind deconvolution)
• Modifications on original Richardson-Modifications on original Richardson-Lucy algorithmLucy algorithm
• More functions in the IDL programMore functions in the IDL program