E£t§acto§ User's guide E. BERTIN Institut d'Astrophysique de Paris
Contents
1. Y centroid in world coordinates 2nd order moment along the major axis image 2nd order moment along the minor axis image 2nd order moment along the major axis world 2nd order moment along the minor axis world Position angle of the major axis counter clockwised 0 0 X axis Position angle of the major axis counter clockwised 0 0 X world axis Isophotal magnitude Isophotal magnitude rms error Corrected isophotal magnitude Corrected isophotal magnitude rms error Fixed aperture magnitude Fixed aperture magnitude rms error WORLD coordinates are computed using informations found in the image FITS header such as the pixel scale 18 MAG_AUTO MAG_AUTO_ERR MAG_BEST MAG_BEST_ERR KRON_RADIUS FLUX_MAX MU_MAX THRESHOLD MU_THRESHOLD BACKGROUND ISOAREA_IMAGE ISOAREA WORLD FWHM_IMAGE FWHM_WORLD ELONGATION ISOn n 0 7 CLASS_STAR FLAGS Automatic aperture magnitude Automatic aperture magnitude rms error MAG_AUTO if there are no neighbours or MAG_ISOCOR otherwise BEST_MAG magnitude rms error Extension of the AUTO aperture in units of A or B Peak surface brightness ADU Peak surface brightness mag arcsec Level of the lowest isophote ADU Level of the lowest isophote mag arcsec Local background ADU Area of lowest isophote pixels Area of lowest isophote arcsec FWHM of profile from a gaussian fit to the core FWHM of profile arcsec Is2. stars i e CLASS_STAR gt 0 5 although I know that most of them are galaxies A This is normal when the SExtractor classifier meets an object too faint to be classifiable there is roughly 50 probability to get a stellarity index greater than 0 5 This is a natural consequence of the fact that the classifier doesn t know a priori the relative fraction of objects which are stars or galaxies There is one solution to this problem one should consider that a faint object might be reliably classified as a point source only if its stellarity index is greater than say 0 9 instead of 0 5 as seen in 2 6 the stellarity index is also a kind of confidence estimate of the classification Q Some saturated stars are classified as galaxies by the neural network despite correct PIXEL_SCALE and SEEING_FWHM parameters in the configuration file A It s normal that saturated stars may be sometimes misclassified because the training of the neural network was done with saturation features that may differ from your image ones As galaxies are generally unsaturated one can easily identify very bright stars according to their saturation flag 3 3 22 Q I get absurd or unaccurate values for FWHM_IMAGE and or FWHM_WORLD A FWHM_IMAGE results from a fit done only on unsaturated pixels of the object Check if the SATUR_LEVEL parameter is set to a proper value Q Strange values are given for magnitudes and the deblending doe
3. 9 You are then nearly sure that the SEEING_FWHM parameter is very close to its optimal value If your image contains stars which are bright enough you might also use the sexseeing SEx Tool The amount of CPU involved in FWHM computations is far from beeing negligible in a crowded field so be careful not to leave FWHM_IMAGE by mistake in the param file if you don t really need it 14 CLASS_STAR o o o o raw gt O 16 18 20 22 24 MAG_BEST Figure 8 CLASS_STAR parameter as a function of MAG_BEST magnitude after adjustment of SEEING_FWHM Interpretation For each classified object the value of CLASS_STAR is an estimate of the con fidence given by the classifier It would be dangerous to interpret this directly as a bayesian probability to be a star since the network training has been done with a synthetic sample hav ing slightly different properties The fig 8 shows objects from the same image as fig 7 in the MAG_BEST CLASS_STAR space note the vanishing constrast between the two classes as magnitude increases NOTE THE STAR GALAXY CLASSIFIER IS STILL EXPERIMENTAL HENCE BE SURE TO CHECK CLASSIFICATION RESULTS BEFORE USING THEM 2 7 Catalog output The catalog is written object by object during the extraction The file name should be specified with CATALOG_NAME It is possible to choose between two formats through the CATALOG_TYPE keyword ASCII or FITS ASCII catalogs contain one object per line in a
4. Fouqu P Gitton J P Kneib V Lebrun and lan Smail for testing and comments Porting to VMS has been made possible thanks to Roderick Johnstone at Cambridge References 1 Beard S M McGillivray H T Thanisch P F 1990 MNRAS 247 311 2 Bertin E 1994 in Science with Astronomical Near Infrared Arrays eds Epchtein N Omont A Burton B and Persi P Kluwer 3 Bertin E Arnouts S 1996 A amp AS in press 4 Bertin E 1996 Th se de Doctorat Universit Paris VI 5 Cotton W D Tody D Pence W D 1995 A amp AS 113 159 6 Da Costa G S 1992 in Astronomical CCD Observing and Reduction Techniques ed How ell S B ASP Conf Series 7 Infante L 1987 A amp A 183 177 8 Irwin M J 1985 MNRAS 214 575 9 Jarvis J J Tyson J A 1981 A J 86 476 20 Figure 9 Check images produced by SExtractor from a 256 x 256 simulated image From left to right top to bottom Original BACKGROUND CONVOLVED OBJECTS SEGMENTATION and APERTURES see text for details The constrast of the BACKGROUND image has been greatly exagerated 10 Kendall M Stuart K 1977 The Advanced Theory of Statistics Vol 1 Charles Griffin amp Co London 11 Kron R G 1980 ApJS 43 305 12 Lutz R K 1979 The Computer Journal 23 262 13 Maddox S J Efstathiou G Sutherland W J 1990 MNRAS 246 433 14 Moffat A F J 1969 A amp A 3 455 15 Odewahn S C Stockwell E B Pennington R L
5. Humphreys R M Zumach W A 1992 AJ 103 318 A FAQ and Troubleshooting Here is a short compilation of answers to Frequently Asked Questions and solutions to some usual problems that a SExtractor user may encounter Don t hesitate to send me a mail in case of trouble bertin iap fr Question Can I do precise photometry with SExrtractor Answer the purpose of SExtractor magnitudes is mainly to obtain photometry of faint sources with the lowest error For bright objects MAG_ISOCOR MAG_AUTO and obviously MAG_BEST mag nitudes do not permit to go beyond 1 in photometric accuracy Moreover they contain a systematic offset compared to true total magnitudes of about 0 05 mag see 2 4 Of course the offset vanishes when the frames are calibrated using the same magnitudes However only by measuring standard stars with MAG_APER aperture magnitudes one will be able to establish very precise zero points down to a few thousandths of magnitude In that case the offset within MAG_ISOCOR MAG_AUTO and MAG_BEST has to be taken into account Q I cannot make the neural network classification work properly on my digitized Schmidt plate images A Unfortunately the present neural classifier makes the assumption that the detector is linear and therefore is not suited to work on photographic images which require a training more specific see Bertin 1995 Q In my faint galaxy survey half of the dimmest objects are classified as
6. PIXEL_SCALE and the seeing Full Width at Hall Maximum SEEING_FWHM both in arcsec In fact the only sensitive parameter is the SEEING_FWHM PIXEL_SCALE ratio You don t need to know precisely the pixel scale as long as you know the seeing FWHM in pixels with a good accuracy 5 10 If not you can run SExtractor a first time on the same image and ask for the FWHM_IMAGE or FWHM_WORLD parameter it is intended to give a good estimate of the FWHM for non saturated bright stars Besides the FWHM can also be used to separate astronomical objects from image defects like cosmic rays fig 7 Note there are several ways to compute the FWHM of a profile They can lead to significantly different estimates because the PSF is almost never perfectly gaussian especially in the wings The true FWHM is generally unusable since there are not enough pixels involved The FWHM_IMAGE parameter is computed by doing a gaussian fit over the upper 80 of the profile It is corrected for undersampling effects down to FWHM lt 1 pixel assuming an underlying gaussian profile It proves to be accurate within a few percents for bright stars This definition serves as the reference for tuning the SEEING FWHM parameter Here is a trick to tune SEEING_FWHM at best pick out a faint ambiguous object in the image what may be a compact galaxy for instance By adjusting SEEING_FWHM you should be able to bring its CLASS_STAR parameter to an intermediate value between 0 1 and 0
7. human readable form whereas FITS catalogs store data as a FITS binary table IEEE format For large fields we recommend to use the FITS option because it produces smaller catalogs However ASCII catalogs are easier to handle when they contain less than say 10 objects FITS binary tables can be read by astronomical packages like IDL IRAF of MIDAS Whichever format used one selects the nature and the name of each parameter listed in the However if an abort occurs at some moment during the working of SExtractor the catalog may not reflect the real extraction status because of buffering 10The FITS binary table format has recently been formally approved by IAU as part of the FITS standard Cotton et al 1995 15 catalog in a parameter file see 3 2 The name of that file has to be supplied with the PARAMETERS NAME keyword 3 Using SExtractor SExtractor is run from your shell with the following syntax sex Picture c configuration filel Parameter Valuetl Parameter2 Value2 The part enclosed within brackets is optional Any Parameter Value statement in the command line overrides the corresponding definition in the configuration file see below 3 1 The Configuration file Each time SExtractor is run it looks for a configuration file If no filename is given in the command line the program takes all its parameters from a file called default sex in the current directory 3 1 1 Format of t
8. string Name of the file containing the neural network weights for star galaxy separation float Threshold to be used for detection keyword Type of threshold value SIGMA number of standard deviations in the background noise MAGNITUDE magnitude per sq arcsec interesting for calibrated images keyword How much SExtractor comments its operations QUIET run silently NORMAL display warnings and limited info concerning the work in progress FULL display a more complete information and the prin cipal parameters of all the objects extracted 3 2 The Parameter file All the parameters that will be listed in the output catalog for every detection should be given in this file The format is ASCII and there must be only one keyword per line The order in which the parameters will be listed in the catalog are the same as that of the keywords in the parameters file Comments are allowed they must begin with a Here is a descriptive list of available parameter keywords NUMBER ISOAREA_PIXEL X_IMAGE Y_IMAGE X WORLD Y_WORLD A_IMAGE B_IMAGE A WORLD B_WORLD THETA_IMAGE THETA WORLD MAG_ISO MAG_ISO_ERR MAG_ISOCOR MAG_ISOCOR_ERR MAG_APER MAG_APER_ERR Identification number Area in pixels within the extraction thresholds isophote X centroid in image coordinates 1 0 center of the first pixel Y centroid in image coordinates 1 0 center of the first pixel X centroid in world coordinates
9. the field is crowded or when the detection threshold is set very low The deblending method adopted in SExtractor is based on multithresholding and works on any kind of object but it is unable to deblend components that are so close that no saddle is present in their profile However as no assumption has to be made on the shape of the objects it is perfectly suited for galaxies and high galactic latitude stellar fields Our method is an attempt to deblend practically any kind of astronomical object Typical problematic cases include patchy extended Se galaxies which have to be considered as single entities and close or interacting pairs of optically faint galaxies which have to be considered as separate objects Basically the multi thresholding algorithm employes a multiple isopho tal analysis technique similar to those in use at the APM and the COSMOS machines Beard McGillivray and Thanish 1991 in a first time each extracted set of connected pixels is rethresh olded at N levels linearly or exponentially spaced between its primary extraction threshold and its peak value This gives us a sort of 2 dimensional model of the light distribution within the object s which is stored in the form of a tree structure fig 3 Then the algorithm goes downwards from the tips of branches to the trunk and decides at each junction whether it shall extract two or more objects or continue its way down To meet the conditions described ear
10. SExtractor 1 0a User s guide E BERTIN Institut d Astrophysique de Paris present address Sterrewacht Leiden PO Box 9513 2300 RA Leiden The Netherlands Contents 1 What is SExtractor 2 Technical overview of the software 2 1 Background estimation o ec cccp ee 2 2 Thresholdine at ohare do B AA da Be de e wk do a 2 37 Deblendine s i A tse bc dd td de ld Bl Boke a 2A sPhotometiy Jet Att Ska Me a des Go ls ke ed AN od i 20 CLEANING Ot da br o dd a tt dd de a ede a 2 6 Star salary separation 2 este as Pets Bel oe ke eu ad 5 XT Catalog output bs ba a le Soe i Pe te ed a 3 Using SExtractor 3 1 The Configuration fle eee 3 1 1 Format of the configuration file o 3 1 2 Parameter listi 4 5 ser Aon rc a Aai ara ee Ss A dr RE A a 3 2 The Parameter file ci na 23 9 4 Atm a MAB eta an RASA RA a 3 9 Flag meaning 4 ze a dee oy bt eda Ba A Re AS 3 4 The Convolution file eee 3 0 Ghe ck image Ans zen ita Aleh a ke Bl A ek AS A FAQ and Troubleshooting B Information for programmers B 1 Add a new parameter B 2 Add a new opiom 4 100 a serene ee ae a AA el B O ala nn amp 12 13 15 16 16 16 16 18 19 19 20 22 1 What is SExtractor SExtractor Source Extractor is a program that builds a catalog of objects from an astronomical image It is particularly oriented towards reduction of large scale galaxy survey data but it also performs wel
11. VOLVE_NORM DEBLEND_MINCONT DEBLEND_NTHRESH DETECTION_TYPE EXTRACT _MINAREA GAIN MAG_GAMMA MAG_ZEROPOINT MEMORY_BUFSIZE MEMORY_PIXSTACK PARAMETER_NAME PHOTOM_APERTURE PHOTOM_KRMIN PHOTOM_KPAR PHOTOM_KRMAX keyword ASCII FITS string keyword NONE BACKGROUND BACKGROUND CONVOLVED OBJECTS SEGMENTATION APERTURES boolean integer float boolean string boolean float integer keyword CCD PHOTO integer float float float integer integer string float float float float Format of output catalog the simplest but space and time consuming FITS format binary table Filename for the check image Type of information to put in the check image normal mode no check image background map background substracted image background substracted CONVOLVE Y objects detected display patches corresponding to pixels attributed convolved image if to each object MAG_APER and MAG_AUTO integration limits If true a cleaning of the catalog is done before being written to disk Number of catalog entries in the stack of objects for cleaning Multiply by 100 to have the memory space occupied in bytes Efficiency of cleaning If true a convolution mask is applied to the data before extraction Name of the file containing the convolution mask If true a normalization of the convolution mas
12. automatic magnitudes Q I can hardly deblend a cluster of small objects even with a very small DEBLEND_MINCONT A The background mesh size BACK_XSIZE and BACK_YSIZE is too large compared to the size of objects to extract see 2 1 Q On my SUN workstation I get a bus error on a very large object with SExtractor On some other workstation with not much memory space I get an error message A Too much memory is required for deblending try to decrease DEBLEND_NTHRESH On SUN a problem with the malloc function sometimes produces a bus error in such occasions B Information for programmers SExtractor is completely written in ANSI C and needs no specific library to be compiled If you re familiar with C you can easily add new functions and parameters to SExtractor Here is a short guide For more details please refer to comments in the source code or send me a mail bertin iap fr 23 B 1 Add a new parameter How SExtractor is storing information about objects it detects After detection Parameters related to the object ex magnitude isophotal area are stored within two kinds of structures The first one concerns data that need information coming from the pixel stack or the image buffer see 2 2 This includes total flux positions etc When CLEANing is on all these informations are kept temporarily in the object stack The object structure is typedef ed objstruct in the include file types h objstructs ca
13. covers the whole frame Our background estimator is a combination of x o clipping and mode estimation similar to the one employed in Stetson s DAOPHOT program see e g Da Costa 1992 Briefly the local background histogram is clipped iteratively until convergence at 30 around its median if o is changed by less than 20 during that process we consider that the field is uncrowded and we simply take the mean of the clipped histogram as a value for the background otherwise we estimate the mode with Mode 2 5 x Median 1 5 x Mean 1 This expression is different from the usual approximation Mode 3 x Median 2 x Mean 2 e g Kendall and Stuart 1977 but was found more accurate with our clipped distributions from the simulations we made Fig 2 shows that the expression of the mode above is considerably less affected by crowding than a simple clipped mean like the one used in FOCAS Jarvis and Tyson 1981 or by Infante 1987 but is 30 noisier That s why we turn back to the mean for uncrowded fields Once the grid is set up a median filter can be applied to it in order to suppress possible local overestimations due to bright stars The resulting background map is then simply a bilinear interpolation between the meshes of the grid The choice of the mesh size is very important BACK_XSIZE and BACK_YSIZE When it is too small the background estimation is affected by the presence of objects and random noise When it i
14. e is done in seven steps estimation of the sky background thresholding deblending filtering of detections photometry classification and finally catalog output Let s now enter a little more into the details of each of these operations in the following keywords in typewriter font refer to parameters from the configuration file or from the parameter file see 3 1 and 3 2 Background Large Image g g Map Frame Butter CE Current Object Output Catalog Figure 1 Global organization of the main SExtractor procedures The pipeline through which all detected objects are processed is enclosed within dashed lines 2 1 Background estimation The value measured at each pixel is a function of the sum of a background signal and light coming from the objects we are interested in To be able to detect the faintest of these objects and also to measure accurately their fluxes we need to have an accurate estimate of the background level in any place of the image a background map Strictly speaking there should be one background map per object that is what would the image look like if that object was absent But here we suppose that most objects do not overlap which is generally the case for high galactic latitude fields To construct its background map SExtractor makes a first pass through the pixel data com puting an estimator for the local background in each mesh of a grid that
15. e to be deblended The dashed vertical line is the theorical limit for unsaturated stars with equal magnitudes In the centroid plot the arrow indicates the direction of the neighbour The simulation assumes a 1 hour exposure with the OCA telescope on a IllaJ plate and Moffat profiles with a seeing FWHM of 3 pixels 2 2 4 Photometry SExtractor has currently the possibility to compute four types of magnitude isophotal corrected isophotal fixed aperture and adaptive aperture For all magnitudes a zero point correction can be applied with the MAG_ZEROPOINT keyword Isophotal magnitudes are computed simply using the detection threshold as the lowest isophote Corrected isophotal magnitudes can be considered as a quick and dirty way for retrieving the fraction of flux lost by isophotal magnitudes If we make the assumption that the intensity profiles of the faint objects recorded on the plate are roughly gaussian because of atmospheric blurring then the fraction 7 feos of the total flux enclosed within a particular isophote reads see Maddox et al 1990 1 At 1 In 1 y 3 a m 3 where A is the area and t the threshold related to this isophote Eq 3 is not analytically invertible but a good approximation to y error lt 107 for y gt 0 4 can be done with the second order polynomial fit A t Her n 1 0 1961 0 7512 G 4 180 180 the THRESHOLD surface brightness magnitude is als
16. ffer margin size Photographic photometry In PHOTO mode DETECT_TYPE PHOTO SExtractor assumes that the response of the detector over the dynamic range of the image is logarithmic This is generally a good approximation for photographic density on deep exposures Photometric procedures described above remain unchanged except that for each pixel we apply first the The buffer margin size is the number of lines that is kept constant except at the end of the frame between the current scan line and the bottom limit of the frame buffer 11 transformation I Ip 107 7 where 7 MAG_GAMMA is the contrast index of the emulsion D the original pixel value from the background substracted image and fo is computed from the magnitude zero point mo 7 0 4m Lo 10 9 8 1n10 8 One advantage of using a density to intensity transformation relative to the local sky background is that it corrects to some extent large scale inhomogeneities in sensitivity see Bertin 1996 for details Errors on magnitude An estimate of the error is computed through is available for each type of magnitude It Am 1 0857 42 m 1 0857 9 where A is the area in pixels over which the total flux F in ADU is summed o the standard deviation of noise in ADU estimated from the background and g the detector gain GAIN parameter in e ADU For corrected isophotal magnitudes a term derived from Eq 4 is quadratically added
17. he configuration file The format is ASCII There must be only one parameter set per line following the form Parameter Value Extra spaces or linefeeds are ignored Comments must begin with a ff and end with a linefeed Values can be of different types strings enclosed between double quotes floats integers keywords or boolean Y y or N n Any missing parameter is set to its default value and a warning message is printed 3 1 2 Parameter list Here is a list of all the parameters known to SExtractor For a detailed description of their meaning please refer to chapter 2 Parameter type keyword Description BACK FLTRXSIZE integer Width in pixels of the background filtering mask BACK FLTRYSIZE integer Height in pixels of the background filtering mask BACK XSIZE integer Width of a background mesh in pixels BACK YSIZE integer Height of a background mesh in pixels BACKPHOTO_THICK integer Thickness in pixels of the background LOCAL annulus BACKPHOTO_TYPE keyword Background used to compute magnitudes GLOBAL taken directly from the background map LOCAL recomputed in a rectangular annulus around the object CATALOG_NAME string Name of the output catalog If the name STDOUT is given and CATALOG_TYPE is ASCII then the catalog will be piped to the standard output stdout 16 CATALOG_TYPE CHECKIMAGE NAME CHECKIMAGE_TYPE CLEAN CLEAN_OBJSTACK CLEAN_PARAM CONVOLVE CONVOLVE_NAME CON
18. image we give it the seeing FHWM as a control parameter Training In order to make the network learn how to distinguish point sources from galaxies we trained it with more than 10 images of stars and galaxies simulated with different conditions of pixel scale seeing and detection limits The PSF used was a Moffat 1969 function with a variable 2 lt 9 lt 4 parameter defining the extent of the wings With this training the network should be able to deal with most ground based telescope linear images with a seeing FWHM comprised between 0 025 and 5 5 The result of the training connection weights was saved to the default nnw ASCII table default nnw is presently the only neural network weights table available Therefore the STARNNW_NAME should be set to default nnw in the configuration file Usage It is very simple to classify stars and galaxies with SExtractor Still for best results with faint objects one needs to do some fine tuning This is the price one has to pay for having We may add other ones in the future 13 FWHM_IMAGE 16 18 20 22 24 MAG_BEST Figure 7 FWHM_IMAGE parameter as a function of MAG_BEST magnitude for objects detected in a deep R band CCD frame Note the horizontal alignment of non saturated stars at FWHM 2 1 pixels as well as the cosmic rays in the lower right part of the plot a high discrimination power Two configuration file parameters are concerned the pixel size
19. is the number of deblending thresholds DEBLEND_NTHRESH A good value is 32 Higher values are generally useless except perhaps for images having an unusually high dynamic range In case of memory problems decreasing the number of thresholds to say 8 or even less may be a solution But then of course a degradation of the deblending performances may occur The second parameter is the contrast parameter DEBLEND_MINCONT As described above a value of 0 005 gives best results Putting it to 0 means that even the faintest local peaks in the profile will be considered as separate objects Putting it to 1 means that no deblending will be authorized The last parameter concerns the kind of scale used for the thresholds If the image comes from photographic material then a linear scale has to be used DETECTION_TYPE PHOTO Otherwise for an image obtained with a linear device like a CCD an exponential scale is more appropriate DETECTION_TYPE CCD 0 4 F Centroid m 19 4 0 2 F XA an i 7 EN v E bA NA 0 2 F js ES v vw Centroid error pixels o T 0 4 F y 0 2 Magnitude J 0 1 l J 0 1 F Magnitude error o T 0 2 F i 4 Separation pixels Figure 4 Centroid and corrected isophotal magnitude errors for a simulated 19 magnitude star blended with a 11 15 19 and 21 mag companion as a function of distance expressed in pixels Lines stop at the left when the objects are too clos
20. k is done before using it Minimum contrast parameter for deblending Number of deblending sub thresholds Type of device that produced the image for linear detectors like CCDs for photographic scans Minimum number of pixels above threshold for detection Gain used for error estimates of CCD magnitudes e7 ADU 7 of the emulsion takes effect only in PHOTO mode Zero point offset to be applied to magnitudes Number of scan lines in the image buffer Multiply by 4 times the frame width to get equivalent memory space in bytes Maximum number of pixels that the pixel stack can contain Multiply by 16 to get equivalent memory space in bytes The name of the file containing the list of parameters that will be computed and put in the catalog for each object Aperture diameter in pixels used by MAG_APER Minimum radius in sigmas for MAG_AUTO magnitudes Kron parameter for MAG_AUTO magnitudes Analysis radius in sigmas for MAG_AUTO magnitudes 17 PIXEL_SCALE SATUR_LEVEL SCAN_ISOAPRATIO SEE ING_FWHM STARNNW_NAME THRESHOLD THRESHOLD_TYPE VERBOSE_TYPE float Pixel size in arcsec for surface brightness parameters FWHM and star galaxy separation only float Pixel value above which an object is considered as sat urated affects the saturation flag and FWHM_IMAGE float Maximum isophotal to aperture object size allowed float FWHM of stellar images in arcsec only for star galaxy separation
21. ks One solution to this problem is to verify for each detection if the object would have been detected provided there were no neighbours This is what the CLEA Ning procedure does Important this error must be considered only as a lower value since it does not take into account the complex uncertainty on the local background estimate Setting GAIN to 0 in the configuration file is equivalent to g 00 12 while detections are made objects are put in a FIFO First In First Out stack When the stack is full objects are examined one by one before being sent to the final catalog SExtractor checks that their mean surface brightness is still higher than the threshold when neighbours are removed The contribution from the wings of neighbours is computed assuming a gaussian extrapolation of their profiles As real profiles have in general broader wings than a pure gaussian it is most of the time necessary to expand their estimated width by a certain factor J clean Practically it is advised to always leave the CLEAN option to Y except for heavily crowded fields or when speed is really critical The CLEAN_PARAM parameter controls the felean factor Typical values range between 1 0 and 2 0 a good compromise is 1 5 Finally the object stack size CLEAN_OBJSTACK should be large enough so that a significant height of the image is considered during cleaning 2 6 Star galaxy separation In most imaging surveys a separation of sta
22. l on moderately crowded star fields Its main features are e Simplicity of usage and configuration e Speed typically 50 kpixel s with a SUN Sparc 20 e Ability to work with very large images up to 65534 x 65534 pixels without being limited by memory e Robust deblending of overlapping extended objects e Possibility to convolve the image on the fly to improve detectability e Neural Network based star galaxy classifier e Flexible catalog output of desired parameters only e Special mode for photographic scans e Modularity of the code that enables one to implement ones own parameters In short the goal in making SExtractor was to find a compromise between refinement in both detection and measurements and computational speed Note the original software was made for Schmidt plate scans only This version is an adaptation which is intended to run on any 2 dimensional astronomical FITS frame coming for example from a CCD or an InfraRed array But for historical reasons many tests shown here have been conducted on photographic material or simulations 2 Technical overview of the software The global working of SExtractor is represented in fig 1 From now on we assume that before using SExtractor For CCD frames flatfielding removal of cosmics and bad columns For IR arrays skysubstraction flatfielding and removal of glitches has been done in order to obtain reliable results The complete analysis of the imag
23. lier the following simple decision criteria are adopted at any junction threshold t any branch will be considered as a separate component if 1 the integrated pixel intensity above t of the branch is greater than a certain fraction 6 of the total intensity of the composite object 2 condition 1 is verified for at least one more branch at the same level Note that ideally condition 1 is both flux and scale invariant However for faint poorly resolved objects the efficiency of the deblending is limited mostly by seeing and sampling From the analysis of both small and extended galaxy images a compromise value for the constrast parameter 6 of 0 005 has proved to be optimum This should normally exclude to separate objects with a difference in magnitude greater than 6 However using a linear scale for the thresholds with photographic material it is possible to go up to 10 mag The same figure applies for an exponential scale with a linear device CCD The outlying pixels with flux lower than the separation thresholds have to be reallocated to the proper components of the merger To do so we have opted for a statistical approach at each faint pixel we compute the contribution which is expected from each sub object using a bivariate gaussian fit to its profile and turn it into a probability for that pixel to belong to the sub object For instance a faint pixel lying halfway between two close bright stars having the same mag
24. matrix 5 7 It is not necessary to enter normalized data just put the CONVOLVE_NORM flag to Y if you want the sum 19 of weights to be 1 3 5 Check image It is often interesting to see what SExtractor is doing with your frame especially when one tries to tune the extraction parameters to obtain the best result When the frame is small enough so that it can fit into memory one has the possibility to ask SExtractor to produce a check image while it processes the data This check image is a FITS image containing information about the extraction There are currently 7 kinds CHECKIMAGE_TYPE of check images e BACKGROUND interpolated background useful to adjust the background parameters like the mesh size e BACKGROUND difference between the image and the interpolated background e CONVOLVED background substracted convolved image Note CONVOLVE must be set to Y e OBJECTS objects detected above the threshold e SEGMENTATION each pixel of the image is assigned a value corresponding to the object it belongs to Very useful to adjust deblending parameters e APERTURES MAG_AUTO elliptical and MAG_APER circular aperture limits are superimposed to the background substracted image MAG_AUTO ellipses surrounding objects flagged as crowded are dashed Fig 9 gives an idea of what we obtain on a crowded field Acknowledgments Many many thanks to S Arnouts J F Claeskens E Copet E Deul P
25. n be grouped in objlists typedef ed objliststruct which can have pointers to pizel lists The second one contains data that can be processed without pixel information in a blind way There s no need to store arrays of such a structure typedef ed obj2struct as its elements can be computed at any time Pixel parameters are computed in two functions preanalyse for basic parameters needed at all stages of processing and analyse for more complex ones In most functions they are generally called obj gt param Blind parameters are only computed just before being written to the catalog file in endobject They are always called outobj2 param Many parameters are computed only if they appear in the param file Modify the source to compute a new parameter Let s suppose you want SExtractor to compute a new parameter myparam and write it to the catalog e First insert myparam in the objstruct or obj2struct type definition in types h Valid types are BYTE short long float and double e Add a parameter name less than 16 characters long the right T_type a pointer amp outobj myparam or outobj2 myparam and a format to param in param h e Insert the portion of code computing the parameter or a call to a function in analyse or endobject in analyse c e Ifthere are dependencies with other parameters add them in updateparamflags catout c e If myparam is a pixel parame
26. n created the program enters its actual extraction phase search for connected sets of pixels above a given threshold For large images the whole data cannot fit into the machine s memory That s why only a fraction of the frame is accessible to SExtractor at any time In the following we shall call that fraction the frame buffer The frame buffer can be seen as a window that has the same width as the full image and moves along line by line Its height in scan lines has to be specified by the user according to the amount of memory available on the machine MEMORY BUFSIZE It can be as small as 4 lines high but while very small frame buffers don t affect source extraction or deblending they certainly slow down the program a bit and above all they represent a handicap for aperture photometry see 2 4 To perform thresholding SExtractor uses a very efficient one pass 8 connectivity algorithm by Lutz 1979 The extraction parameters definable by the user are the threshold expressed in number of background noise standard deviations or in mag arcsec if THRESHOLD_TYPE MAGNITUDE and the minimum number of connected pixels required EXTRACT MINAREA Typical values of 1 0 to 2 0 for THRESHOLD and 5 for EXTRACT_MINAREA generally give the best results A convolution can be applied to the image before extraction by setting CONVOLVE to Y The convolution mask CONVOLVE_NAME is supplied by the user see 3 4 and can be of any size wa
27. nitude will be appended to one of these with equal probabilities One big advantage Density Area Figure 3 A schematic diagram of the method used to deblend a composite object The areal profile of the object smooth curve can be described in a tree structured way thick lines The decision to regard or not a branch as a distinct object is determined according to its relative integrated intensity tinted area In that case above the original object shall split into two components A and B Remaining pixels are assigned to their most credible progenitors afterwards of this technique is that the morphology of any object is completely defined simply through its list of pixels To test the effects of deblending on photometry and astrometry measurements we made several simulations of photographic images of double stars with different separations and magnitudes under typical observational conditions fig 4 It is obvious that multiple isophotal techniques fail when there is no saddle point present in profiles i e for distance between stars lt 20 in the case of gaussian images We measured a magnitude error lt 0 2 mag and a shift of the centroid lt 0 4 pixels for the fainter star in the very worst cases but no other systematic effects were noticeable The user can control the multi thresholding operation through 3 parameters The first one
28. o affected by MAG_ZEROPOINT A total magnitude Mmo estimate is then Mot Miso 2 5 log 7 5 Clearly this cheap correction works best with stars and although it is shown to give tolerably accurate results with most disk galaxies it fails with ellipticals because of the broader wings of their profiles Therefore we recommend to apply it only as a better than nothing substitute to aperture magnitude in crowded cases Fixed aperture magnitudes estimate the flux above the background within a circular aper ture The diameter of the aperture in pixels PHOTOM_APERTURE is supplied by the user in fact it does not need to be an integer since each normal pixel is subdivided in 5 x 5 subpixels before measuring the flux within the aperture Automatic aperture magnitudes are intended to give the most precise estimate of total magnitudes at least for galaxies Our automatic aperture photometry routine is inspired by Kron s first moment algorithm 1980 1 We define an elliptical aperture whose elongation e and position angle O are defined by second order moments of the object s light distribution The ellipse is scaled to Rmaz Ciso typically 60 so which corresponds roughly to 2 isophotal radii 2 Within this aperture we compute the first moment 2e a Xlr Kron 1980 and Infante 1987 have shown that for stars and galaxy profiles convolved with gaussian seeing gt 90 of the flux is expected to lie
29. ophotal weighted elongation equivalent to A IMAGE B_IMAGE Isophotal area for subsequent profile classification pixels See Bertin 1994 Stellarity index given by the neural network output 0 0 Galaxy and 1 0 Star Extraction flags see below 3 3 Flag meaning In order to be easily readable by most programs SExtractor s FLAGS parameter contains coded in decimal all the extraction flags as a sum of powers of 2 1 The object has neighbours bright and close enough to bias significantly Am 0 1 MAG_AUTO 2 The object was originally blended with another one 4 At least one pixel of the object is saturated or very close to 8 The object is truncated too close to an image boundary 18 Object s aperture data are incomplete or corrupted 32 OQObject s isophotal data are incomplete or corrupted 64 A memory overflow occured during deblending 128 A memory overflow occured during extraction For example an object close to an image border may have FLAGS 16 and perhaps FLAGS 8 16 32 56 3 4 The Convolution file The convolution file contains the elements of the convolution matrix applied to the data before extraction the CONVOLVE preference flag must have been set to Y before The format is ASCII Here is an example of a 3 x 4 matrix CONV a comment 1 5 3 46 1 8e 02 4 01 58 2e 1 20 2 2 2 1 11 The string CONV in the first line specifies that the file is a convolution
30. rs and galaxies is required This is traditionally a tricky job the seeing and the limited signal noise both contribute to bluring the distinction between faint point sources and extended objects We have chosen to confide this classification task to a neural network The superior performances of neural networks in the classification of photographic images have already been demonstrated Odewahn et al 1992 Bertin 1994 Even with linear images they can lead to significant improvement over traditional techniques espe cially concerning their robustness when dealing with overlapped objects We will now present briefly the classifier for a more complete description please refer to Bertin and Arnouts 1996 Principle First of all the goal of the neural classifier is to provide an optimal transformation from the parameter space defined by a set of observables describing the objects to the space of classes Here the space of classes is one dimensional and from now on we shall refer to it through the stellarity index CLASS_STAR which by definition will equal 0 for galaxies and 1 for stars The input parameters are simply 8 isophotal areas and the peak intensity plus one control parameter It can be shown in pratice that raw parameters such as those do provide a very good description of object profiles and particularly of their fuzziness see Bertin 1996 for more details The network needs to know what is the instrinsic fuzziness of the
31. s not work well on my CCD frame A Check if the DETECTION_TYPE is not set to PHOTO by mistake Q I don t understand how the WORLD parameters are computed A WORLD parameters are computed using astrometric information found in the FITS header of the image file like position offset step per pixel rotation angle ISOAREA WORLD and FWHM_WORLD parameters are a little bit special if PIXEL_SCALE is set to 0 in the config file they are computed using a pixel size deduced from FITS header informations assuming these FITS informations are expressed in degrees Otherwise the PIXEL_SCALE value in arcsec per pixel is used Q The noise estimate given for the background becomes higher as the mesh size increases A Perfectly normal if the background is far from beeing uniform any variation of the background within a mesh can be considered as low frequency noise Q On an APERTURES check image I get one circle and one ellipse around each object A It happens because you selected both MAG_APER and MAG_AUTO or MAG_BEST in the parameter file The circle has a diameter equal to PHOTOM_APERTURE Q On an APERTURES check image ellipses are flattened in the y direction for the largest objects and the MAG_AUTO are too high flux lost A The number of lines in MEMORY_BUFSIZE is too small compared to the size of the largest objects in the frame anyway such objects are flagged as corrupted from the point of view of
32. s too large it cannot reproduce the small scale variations of the background Therefore a good compromise has to be found by the user Typically for reasonably sampled images a width of 32 to 256 pixels works fine The user has again some control over the background map by specifying the size of the median filter BACK_FLTRXSIZE and BACK_FLTRYSIZE A width and height of 1 means that no filtering will be applied to the background grid Usually a size of 3 x 3 is enough but it may be necessary to use larger dimensions especially to compensate in part for small background mesh sizes The background estimation operation can take a considerable time on the largest images e g half an hour for a 32000 x 32000 frame on an HP 755 workstation Obviously in some very unfavorable cases like small meshes falling on bright stars it leads to totally unaccurate results SExtractor offers the possibility to have rectangular background meshes but it is advised to use square ones except in some very special cases rapidly varying background in one direction for ex Clipped Mode ADU o T 0 5 10 15 20 25 30 Clipped Mean ADU Figure 2 Simulations of 32 x 32 pixels background meshes polluted by random gaussian profiles The true background lies at 0 ADU While being slightly noisier the clipped Mode gives a more robust estimate than a clipped Mean in crowded regions 2 2 Thresholding Once the background map has bee
33. tch out for processing time It has been shown that an optimum detection of faint point sources is achievable by convolving the data with a gaussian whose FWHM is roughly equal to the seeing Irwin 1985 Of course it is also possible to use more specific convolution masks for instance to detect structures at given scales Note that convolution affects only detection and deblending However it has for instance little effect on aperture magnitudes Until all the pixels above the threshold of an object have been scanned they are stored in a pizel stack This stack is shared with pixels from all the other unfinished objects The stack size MEMORY_PIXSTACK is set by the user It should be as large as possible depending on the memory available on your machine Indeed when SExtractor has no more room to store incoming pixels it sacrifices the largest object extracted so far to gain memory space It will in most cases be a very bright star or an artefact such as one of the strips often found at the edges of images Sacrificed objects are nevertheless included in the output catalog Of course they are flagged accordingly because their parameters have been corrupted and no deblending procedure is applied to them 2 3 Deblending Each time an object extraction is completed the connected set of pixels passes through a sort of filter that tries to split it into eventual overlapping components This case appears more frequently when
34. ter you can define its behaviour during the CLEANing process in mergeobject clean c That s all B 2 Add a new option Adding a new option to SExtractor is also quite easy let s see how to add the option myoption e First insert myoption in the prefstruct type definition in types h Valid types are int double char strings or enum e Add an option name a P_type the pointer amp prefs myoption imin and imax limits if it s an int dmin and dmax limits if it s a double and a list of keywords if it s an enum to key in prefs h None of the keywords or the option name should have more than 15 characters 24 e Add a FITS keyword SEXMYOPT for instance with no more than 8 characters a T type the pointer amp prefs myoption a format and a H_type to hparam in param h Insert the corresponding line in the FITS header template of initfitscat catout c e If necessary edit useprefs in prefs c to take into account informations taken from the image You can now use the prefs myoption variable to do whatever you want in SExtractor from the configuration file 25
35. to take into account the error on the correction itself o f In PHOTO mode things are slightly more complexe Making the assumption that plate noise is the major contributor to photometric errors and that it is roughly constant in density we can o1n10 gt Ez y Am 1 0857 _ _ 2 10 Yay 1 5 y where I x y is the contribution of pixel x y to the total flux Eq 7 The GAIN is ignored in PHOTO mode write Background is the last point relative to photometry The assumption made in 2 1 that the local background associated to an object can be interpolated from the global background map is no longer valid in crowded regions An example is a globular cluster superimposed on a bulge of galaxy SExtractor offers the possibility to estimate locally the background used to compute magnitudes When this option is switched on BACKPHOTO_TYPE LOCAL instead of GLOBAL the photometric background is estimated within a rectangular annulus around the isophotal limits of the object The thickness of the annulus in pixels can be specified by the user with BACKPHOTO_SIZE 24 is a typical value 2 5 CLEANIing When using low thresholds spurious detections are often created at the wings of objects with shallow profiles for example elliptical galaxies It comes from the fact that the background is locally higher there leading to a lower relative threshold and thus a higher detection rate of noise pea
36. urvey plate image Geometrical constraints When measuring aperture magnitudes all the pixels of an object plus many others around it must be accessible in memory That is they must lie within the upper and lower boundaries of the frame buffer So we have introduced an isophotal to aperture ratio p parameter SCAN_ISOAPRATIO that enables the user to control the geometrical 10 Isophotal Automatic Aperture Corrected Isophotal E Measured mag True mag True total magnitude Figure 5 Flux lost expressed as a mean magnitude difference with different faint object pho tometry techniques as a function of total magnitude see text Only isolated galaxies no blends of the simulations have been considered constraints of aperture photometry If p is less than the ratio A A ele then the object is flagged as aperture truncated flag 16 in 3 3 It is easy to show that it gives also an optimum buffer margin size equal to ne where h is the frame buffer size in pixels see fig 6 A secure value for SCAN_ISOAPRATIO is 0 6 Large Image Frame Buffer BUFFER SIZE OBJECT SIZE Current scan line MARGIN SIZE Scan direction Figure 6 Geometrical constraints concerning aperture magnitudes and frame buffer size Aper ture magnitudes are computed if and only if the ratio object size is less than a certain factor T SiZ SCAN_ISOAPRATIO specified by user SCAN ISOAPRATIO also determines the bu
37. within a circular aperture of radius kr if k 2 almost independently of their magnitude This picture remains unchanged if we consider an ellipse with ekr and kr e as principal axes k 2 defines a sort of balance between systematic and random errors By choosing a larger k 2 5 the mean fraction of flux lost drops from about 10 to 6 When Signal to Noise is low it may appear that an erroneously small aperture is taken by the algorithm That s why we have to bound the smallest accessible aperture to Rmin typically Rmin 30iso Through the configuration file the user has full control over the 3 parameters k Riaz and Rmin respectively PHOTO_KSIG PHOTO_KRMAX and PHOTO_KRMIN Aperture magnitudes are unfortunately sensitive to crowding Therefore we suggest to replace the aperture magnitude by the corrected isophotal one when an object is too close from its neighbours 2 isophotal radii for instance This is done automatically when using the MAG_BEST magnitude MAG_BEST MAG_AUTO when it is sure that no neighbour can bias MAG_AUTO by more than 10 or MAG_BEST MAG_ISOCOR otherwise Experience shows that the MAG_ISOCOR and MAG_AUTO magnitude loose about the same fraction of flux on most images around 0 06 for default extraction parameters Figure 5 shows the mean loss of flux measured with isophotal threshold 24 4 mag arsec corrected isophotal and automatic aperture photometries for simulated galaxy By on a typical Schmidt s
Download Pdf Manuals
Related Search