TN 0034: Some Notes on Using the New Version of Sextractor to Generate Star Lists From TASS Images

Author: Glenn Gombert
Date: 970720
Revision: #0 970720
Key Words: photometry, techniques, computation

                                                      Introduction
On June second Emmanuel Bertin released a new version of his “SExtractor” (or Source Extractor) program which includes a number of new features which should prove quite useful in the reduction of TASS images to Star-Lists. This technical note details some of these new features and how they might be used to improve TASS data reduction.

Shown below are some of the new features (from the Readme file) of the June, 1997 release of SExtractor:

(1). Enhanced support for WCS coordinates. This means that you can get the alpha and delta for objects detected on FITS images which comply with the WCS standard (see:
   http:// www.cv.nrao.edu/fits/documents/wcs/wcs.html)

(2). 2D positional error estimate through ERRA_xxx, ERRB_xxx and ERRTHETA_xxx. Note: these estimates are exact for unconvolved, white-noised images only.

(3). Linear photometric parameters (FLUX_xxx and FLUXERR_xxx in addition to MAG_xxx and MAGER_xxx).

(4). Better outputs: units and vectors in FITS, (limited) backward compatibility with the old V1.0 output format (CATALOG_TYPE FITS_1.0), new detailed ASCII output mode (CATALOG_TYPE ASCII_HEAD).

(5). Handling of gigantic images (up to 2e9x2e9 pixels on 64bit machines).

(6). Online decompression of compressed DeNIS images.

(7). New "vignet" vector output: for instance VIGNET(8,5) in the .param file will write in the output catalog a small 8x5 pixels array around each detection center.

(8). There is now a new "DETECT_IMAGE" parameter which requires a string as argument. If it set to "SAME", it means that the input image will be used for both detection and measurement, as in the old version. If you give it an mage  filename (provided that the dimensions of this image are the same as those of the input image), it will be used as a kind of ``detection template'': all sources will be detected and deblended on this image, and will serve as reference for the measurements done on the regular image.

(9). The old THRESHOLD parameter has been now split in two: DETECT_THRESH and ANALYSIS_THRESH. DETECT_THRESH is the threshold of detection, just as before. You may express it in units of sigma of the background noise, or in surface brightness units, associated to a magnitude Zero-Point of the detection frame.

       Ex: "DETECT_THRESH 3.0" will set the threshold at 3.0 sigma,
      "DETECT_THRESH 22.0,25.4" will set the threshold at 2 mag.arcse-for a detection-image whose mag-ZP is 25.4 mag. (note: the "PIXEL_SCALE" should have been set to the right value). In this way, the THRESHOLD_TYPE parameter has been eliminated. ANALYSIS_THRESH is a threshold in the measurement-frame (it can be the same as the detection-frame). It is used only for estimators that need accurate information about the object profile in the measurement frame star/galaxy separation and FWHM.

(10). Similarly, the 3 Kron parameters PHOTOM_K... have been replaced by a common PHOT_AUTOPARAMS which requires 2 arguments, equivalent resp. to the former PHOTOM_KPAR and PHOTOM_KMINSIG. PHOTOM_KSIG is now set definitely to 6.0 in the code. This is because the new MAG_AUTO apertures are much more robust to crowding effects than before.

(11). Also, for aesthetic reasons, PHOTOM_APERTURE has become PHOT_APERTURES. There is an S at ApertureS, because the software can now photometer each source simultaneously through up to 4 apertures (see default.param). In default.param, MAG_APER is equivalent t MAG_APER1, and MAGERR_APER is equivalent to MAGERR_APER1.

(12). The MAG_AUTO magnitudes are computed in a slightly different way than before, leading to a gain of about 20% in photometric precision in most extragalactic fields, especially in clusters. Obviously MAG_BEST benefits from this.
 

                               Application to TASS Image Reduction:

 Of the new features described above several of them have greatly enhanced the use of “Sextractor” for the reduction of TASS images.

*    The enhanced support of World Coordinate System (WCS) parameters corrected some problems that I ran into when reducing of some data taken here in Dayton for the June AAS meeting. The astrometric errors (in RA) that I found were 2-3 arcseconds, the errors in Declination were typically larger than this (10 arcseconds or more). This let to only four matched (between V&I band data) for the data that was posted to the TASS Home page for the night of April 8/9. When I ran the same raw data files through the new version of “Sextractor” I found that there was 471 matches of the same data. This indicates that the problems that I had encountered before in the previous version of Sextractor mapping WCS parameters to RA and DeC had been fixed in the new release.

*     The new photomertic functions of SExtractor should be of significant help improving the photometric reduction of TASS Images. One new function (#4 above) allows multiple fixed apertures to be calculated at any one time. Also the “MAG_BEST” photometric estimate has been enhanced to provide a much better estimate of the instrumental stellar magnitude than the previous version.

Shown below is a plot of fixed aperture size vs magnitude for a star that was detected in a recent TASS image. The “MAG_BEST” estimate of this star (from SExtractor) is 10.8516 magnitude showing how the flux from the star being measured is recovered with SExtractor’s “adaptive-aperture” magnitude algorithm. This is even more flux recovered that using a large (17 pixel) aperture size shown in the plot below:
 
 

 

    The automatic aperture routine was inspired by Kron’s “first moment” algorithm (Kron R.G. 1980 ApJS 43,305). A description (from the Sextractor Manual) of how this algorithm works is shown below:


 

*  The ability to save the area around each star that is detected in an image should prove useful when storing TASS images in any type of a database. This allows a circular area around each star to be written out to a FITS format table so that this data is not lost if raw images are not saved to disk.

    Arne suggested this option in one of his posts recently concerning TASS Database structures:
 “As far as keeping a pixel archive:  George Turner and I (plus others) have maintained since the beginning that robotic telescopes should extract their data and then throw away the images.  You gain factors of 100-1000x compression. Test your extraction software carefully,and then trust it. The few times you need something less than 1sigma will be extremely rare.  You can always save postage stamps around those objects of interest, or objects that do not appear in the master star list, if you are concerned about losing information.”
 

    An example of the star-list that is generate (with this option enabled is shown blow) for several objects that are detected in the image "g0483967.fts":

x-pos        y-pos    ra                        dec                        mag     flags
109.52      7.49  -1.052273e+000   1.998669e+002  14.0595   000     -65.2861    -46.4941    -49.7139    -24.9443    -100.184      21.542     28.3467     115.139    -4.07813    -59.3057     46.3691     349.186     239.991     34.7861     37.5723    -44.8057     294.023     314.842     121.649    -61.5518    -83.9805     28.8613     59.6924    -39.4873     15.3252

   x-pos        y-pos   ra                        dec                    mag          flags
   474.93      9.35   3.434709e-001   1.998740e+002  11.6518    000      244.931     618.757     538.593     362.438     82.2939     542.822     1596.65     1726.49     896.339     374.196     836.724     2260.56      3610.4     2102.25     652.107     558.637     1554.47     2824.32     2546.17      836.03     326.557     806.395     1044.24      1058.1     515.962

   x-pos        y-pos   ra                        dec                      mag                   flags
   759.01     12.18   1.428556e+000   1.998848e+002  13.2609 000      64.3066     195.654     110.988      108.31     7.61914     114.994     332.347     595.687     247.016     -93.668     171.673     529.031     916.378     587.714     209.038     40.3447     269.709     483.063     638.405     225.737      101.01      102.38      211.74      219.09      106.43

===============================================================
 

*     To detect the effects of crowding on TASS images the “status flags” parameter can be used to detect when a faint star might influence the photometric measurement the star that is being measured. A definition of the SExtractor “flag” parameters is detailed below:

    ( 1) Flag = 1, The object has neighbors, bright and close enough to enough to bias significantly (delta m ~ 0.1)

    (2) Flag = 2, The object was originally blended with another one.

    (3) Flag = 4, At least one pixel of the object is saturated (or very close to).

    (4) Flag = 8, The object is truncated (too close to an image boundary.

    ( 5) Flag = 16, Object’s aperture data is incomplete or corrupt.

    (6) Flag = 32, Object’s isophotoal data is incomplete or corrupt.

    (7) Flag = 64, A memory overflow occurred during deblending.

    (8) Flag = 128 A memory overflow occurred during extraction.
 

        As an example of how this might be useful, shown below is two lines from the star-list that was generated from image "go483967.fts":

      x-pos      y-pos  ra                         dec                     magnitude      flags
    492.74     16.90   1.878284e+000   6.074249e-002  13.1287         002
    492.09     21.48   1.875801e+000   7.821706e-002  12.8884         002
 

        The "flags" parameters indicates that these two objects had been origianlly "blended" together and had been "separated" from each other by the star-extraction process. The small segment of the image below shows these two objects from the original image:

                                                        
 
 

    These flags could be used to tell which measurements one might wish to excluse from a data set (to determine specific conditions upon which a mesurement would not want to be included). (i.e. such as some of the pixel data was saturated, etc. )
 

* The ability to generate "check-images" from SExtractor to tell what kind of performance is being experienced by the Extraction program:

    (1).  Background: interpolated background: useful to adjust the background parameters like mesh-size.

    (2). -Background: difference between the image and interpolated background.

    (3). Convolved: background -subtracted convolved image. Note: CONVOLVE must be set to Y in the parameter file.

    (4). OBJECTS: objects detected above the thrshold.

    (5). Segmentation: each pixel of the image is assigned a value corresponding to the object it belongs to. Very useful to adjust deblending parameters.

    (6). Apertures: MAG_AUTO and MAG_APER circular apeture limits are superimposed to the background-subtracted image. MAG_AUTO ellipses surrounding objects flagged as "crowded" are dashed.

                    Shown below is the image from TN#32 "go483967.fts":
 
 
                           

    Shown below is the "BACKGROUND" plot from the above image that Michael Richmond used in his TN#32.

                             
 

                An example of the "Apertures" plot is also shown for the same image:

                             
 

    The objects that are detected in the image can be seen and the overall operation of the extraction process can be "fine-tuned" using the various "check-image" options of SExtractor.
 
 

                                                Conclusions:

    The above TechNote summarizes some of the more useful featuers of the SExtractor program for reducing TASS Images. A GUI interface for SExtractor is underway for Window95/NT and should be available soon. I will try to add to this TechNot in the future as I discover new features of SExtractor that might be of interest to the TASS effort.