Download ANDRIL package The algorithm has been integrated within SolarSoft
ANDRIL is the iterative deconvolution algorithm based on maximum likelihood approach. The alhorithm is written in the IDL programming language. It has been designed for advanced processing of images obtained by the Soft X-ray Telescope (SXT) on Yohkoh.ANDRIL allows to remove numerically the image blurring due to the instrument point spread function (PSF) and allows to increase the image resolution. It has been used for analysis of detailed morphology and physical conditions in the plasma of flaring structures. (1997, 1999)
Before calling the deconvolution programme, one has to prepare the data. However, we are working on this and will notify registered users when it will be upgraded and integrated.(now obsolate). The program handles at present square SXT images or subimages taken with the full resolution.
The following data preparation steps are necessary before running the programme (pre SolarSoft):
1) .run yodat - and process required image(s) 2) run data preparation sxt_prep,index,data ,index_out, data_out,/float 3) determine "dark current": dark=float(sxt_decomp(data))-data_out 4) call deconvolution for particular image "i" f. i.
andril_sxt, nfil, x0 , y0 , data, dark, decon
The program calls a number of routines described at the bottom of this page. The package assumes that the uncertainty files: ddn0.sav, ddn1.sav, ddn2.sav, ddn3.sav,ddn4.sav, ddn5.sav, ddn6.sav are located in the working directory. You may change appropriate line in the (Andril_SXT.pro in order to redirect the calling (cf. program line: 211 )
The procedure accepts the following keywords: B - is the pedestal background ( b =~10 ) which damps the noise most efficiently (see Paper II) DIMPSF - is the size of the PSF array as expressed in SXT pixels. Default is 7 (for 7x7 PSF window size) DIV - is the superresolution factor (default DIV=5) DN - allows to write the results in DN's/s MODE - is: 0 for classical Richardson-Lucy (RL), 1 for accelerated RL N, T - are the parameters regulating the U - smoothing function (default N=5, T=0.2) (see Paper I ) NITER - is the termination step for the iteration (default NITER=50 except for optical filter for which NITER=10) SILENT - allows to run deconvolution without access to Xterminal (doesn't show the windows for the plots: U-Function, correction matrix, abs(obs-cal)/unc, ratio observed/calculated and deconvolved image) THR - Signal/Noise threshold used for detecting significant pixels, used for calculations of normalized Chi^2 (default THR=5)
The procedure accepts the following keywords:
The results are written to the following files: andril.out - text file with the convergence results "andril.sav or andrilm.sav" - deconvolution results are written to the actual directory. Andril.sav - contains results in photons, andrilm.sav - normalized results in data number/subpixel/sec. This later file is created with the /DN key applied. The meaning of variables in the andril.sav and andrilm.sav files is explained in the appropriate info string present in the save file. Below, we repeat these descriptions: andril.sav
The results are written to the following files: andril.out - text file with the convergence results "andril.sav or andrilm.sav" - deconvolution results are written to the actual directory. Andril.sav - contains results in photons, andrilm.sav - normalized results in data number/subpixel/sec. This later file is created with the /DN key applied. The meaning of variables in the andril.sav and andrilm.sav files is explained in the appropriate info string present in the save file. Below, we repeat these descriptions:
andril.sav
BB - value of "background" pedestal (see Paper II) CALC - matrix of calculated fluxes (photons) CHITH - threshold for S/N (5 default) DARK - array of the dark current used DIMPSF - size of the PSF square array used in deconvolution (input pixels) DIV - over-resolution factor (usually 5) EDGE - 0 if no edges added in calculations, 1 if extra edge IMAGE - input image for deconvolution in photons ITER - iteration limit (or iteration step where convergence stops) ' LIKEL - vector of log L(ikelihood) convergence MODE - mode used in the calculations (0 - classical RL, 1 - accelerated N - scalar defining the smoothing properties (see Paper I ) NEW - THE RESULT of DECONVOLUTION (photons, total counts preserved) PSF - the PSF array used for deconvolution T - scalar defining the smoothing properties (see Paper I ) THR - threshold level for S/N ratio (5 is for 20% accuracy) UNCER - matrix of combined total uncertainties in photons
andrilm.sav BB - value of "background pedestal" in DN's(see Paper II) CALC- matrix of calculated fluxes DNs CHITH - vector of normalized chi^2 convergence (pixels with important signal S/N>5), if no points above threshold are present then the value is calculated using all pixels DARK - array of the dark current used in DNs/s DATEK, FRAMEK - date & frame No in the flare sequence DIMPSF - size of the PSF square array used in deconvolution (input pixels) DIV - over-resolution factor (usually 5) EDGE - 0 if no edges added in calculations, 1 if extra edge IMAGE - input image for deconvolution in DNs/s ITER - iteration limit (or iteration step where convergence stops) LIKEL - vector of log L(ikelihood) convergence MODE - mode used in the calculations (0 - classical RL, 1 - accelerated N - scalar defining the smoothing properties (see Paper I ) NEWNORM - THE RESULT of DECONVOLUTION (DN's/s, total counts preserved) NFIL - filter used PSF - the PSF array used for deconvolution STIMEK - string with time UT T - scalar defining the smoothing properties (see Paper I ) TIMEK - time, decimal hours THR - threshold level for S/N ratio (5 is for 20% accuracy) UNCER - matrix of combined total uncertainties DNs/s X0,Y0 - coordinates of lower left corner of the SXT FOV You may found the schematic diagram of andril_SXT programme here. Inside the downloadable package, there are the following files: adril.sxt.pro - MAIN ROUTINE; sets-up deconvolution of SXT images from full-resolution data adrlb.pro - ANDRIL ENGINE calctst.pro - rebin back the convolved array of intensities into the array of observed SXT pixels from convolved array "con" chisq.pro - calculates normalized chisqr parameter for two arrays measured and calculated when array of uncertatinties is known convolve.pro - convoles the image with Point Spread Function (PSF) data.sav - data file ddn0.sav - data file ddn1.sav- data file ddn2.sav- data file ddn3.sav- data file ddn4.sav- data file ddn5.sav- data file ddn6.sav- data file diwalim1.pro - calculates the weight factor tensor drlb.pro - general ANRIL routine for user specified image unsertainty,PSF etc sets-up deconvolution of PSF from images image_c.pro - overlay an image and a contour plot nlikely.pro - calculates value of the 'normalized' natural LOG of the likelihood function for given arrays of the observed and calculated data. It is IMPORTANT that obs and calc are in photons (or statistical events) in order that the likelihood be meaningfull parabola.pro - calculates selected parameters of parabola positivi.pro - takes unconstrained x (usually an image), and map it uniquely and smoothly into positive values rl_decob.pro - checks if input data for deconvolution are correct matrices ( square and odd/even where nesessary) and executes DRLB for algorithm version "with pedestal B" sxt_comp.pro - compress an image from 12 bits to 8 bits sxt_deco.pro - decompress an SXT image, optionally return decompression uncertainty. Returns an image of the SXT Point-Spread-Function ufun.pro - damping function for new likelihood function following Richard.L White
andrilm.sav
BB - value of "background pedestal" in DN's(see Paper II) CALC- matrix of calculated fluxes DNs CHITH - vector of normalized chi^2 convergence (pixels with important signal S/N>5), if no points above threshold are present then the value is calculated using all pixels DARK - array of the dark current used in DNs/s DATEK, FRAMEK - date & frame No in the flare sequence DIMPSF - size of the PSF square array used in deconvolution (input pixels) DIV - over-resolution factor (usually 5) EDGE - 0 if no edges added in calculations, 1 if extra edge IMAGE - input image for deconvolution in DNs/s ITER - iteration limit (or iteration step where convergence stops) LIKEL - vector of log L(ikelihood) convergence MODE - mode used in the calculations (0 - classical RL, 1 - accelerated N - scalar defining the smoothing properties (see Paper I ) NEWNORM - THE RESULT of DECONVOLUTION (DN's/s, total counts preserved) NFIL - filter used PSF - the PSF array used for deconvolution STIMEK - string with time UT T - scalar defining the smoothing properties (see Paper I ) TIMEK - time, decimal hours THR - threshold level for S/N ratio (5 is for 20% accuracy) UNCER - matrix of combined total uncertainties DNs/s X0,Y0 - coordinates of lower left corner of the SXT FOV