Algorithm 2 -- LCSB Processor

Algorithm 2 is designed to detect faint central surface brightness galaxies. This memo describes the results of algor 2 on the Hercules 5 repeat scans.

Images

The following images are the brightest of the LCSB candidates (snr > 5). The list includes real galaxies, false galaxy detections (either faint stars or simply noise bumps) and unknown object (too faint to call).

• Scan 047: lcsb srcs 1 - 3
  
• Scan 047: lcsb srcs 4 - 4
  
• Scan 048: lcsb srcs 1 - 3
  
• Scan 048: lcsb srcs 4 - 6
  
• Scan 048: lcsb srcs 7 - 9
  
• Scan 048: lcsb srcs 10 - 10
  
• Scan 049: lcsb srcs 1 - 3
  
• Scan 049: lcsb srcs 4 to 5
  
• Scan 050: lcsb srcs 1 - 3
  
• Scan 050: lcsb srcs 4 - 6
  
• Scan 050: lcsb srcs 7 - 9
  
• Scan 050: lcsb srcs 10 - 10
  
• Scan 051: lcsb srcs 1 - 3
  
• Scan 051: lcsb srcs 4 - 6
  
• Scan 051: lcsb srcs 7 - 9
  

LCSB galaxies are typically too faint to employ the 'normal galaxy' alorithms to separate stars from extended sources. We do, however, take advantage of the unique property of LCSBs -- they "block up" more efficiently than stars. Our primary discrimanent is to measure the SNR of the source after it has been blocked up (2 X 2, 4 X 4 and 8 X 8). For a faint star (i.e., the only thing left in the coadd after algorithm 1 has processed the coadd) will have a SNR that is small, while a comparably bright galaxy will have a larger SNR (because of course it is extended). We use variations on this theme to further refine the star-galaxy separation.

One kind of false galaxy that is very troublesome to any kind of low surface brightness galaxy detection is not a star, but rather a set of noise bumps (typically more than one, localized to a small area) that mimics a real galaxy. There are many different flavors of this kind of "background' enhancement false detection, including fuzz near bright galaxies -- a particular problem within the Hercules core because of the giant spirals and ellipticals and high density of bright galaxies. There is not much that can be done about these false detections, other than to use a flux filter since they typically are very faint (well beyond the level-1 spec) and often have a low SNR if measured using the "super" coadd (J+H+K). Another nuisance, meteor streaks, are efficiently detected with algorithm 2. A "human-in-the-loop" can easily filter these objects from the database. The plots below show the results.

Parameter Definitions

The following plots/figures contain information with regard to the various parameters computed for LCSB objects. Here is a brief glossary of the nomenclature and terms.

J10, H10 and K10

Fixed circular JHK magnitudes corresponding to radius=10 pixels.

Jc, Hc, Kc

JHK mags as derived using the "flux-growth" algorithm. The "flux-growth" algorithem works in general as follows:


The object flux is computed by integrating either circular or elliptical apertures that are sequentially increased in size until the flux growth "converges" as measured by either of two indicators:


(1) the change in flux between adjacent apertures is less than 0.5%; this criterion is design to maximize the integrated flux of the object;


(2) the change in mean surface brightness between adjacent aperture-annuli is less than (-3 * sigma), where sigma is the standard deviation of the background counts in the annulus corresponding to a radius = 15 pixels; this criterion is designed to minimized confusion from stars and from background gradients (e.g., nearby extended sources).

JHK: CSB and C5SB

Central surface brightness as measured from the peak pixel and central surface brightness as measured using a circular aperture radius=5.


JHK: BSNR

The peak pixel SNR as measured in the blocked/smoothed JHK images. The noise corresponds to the background noise of the blocked/smoothed images. LCSB candidates that have BSNR = 0 correspond to non-detections in at least one of the bands.


JHK: TSNR

The integrated SNR using circular aperture radius = 10 with the inner radius=2 pixel blanked. This parameter is designed to minimize the SNR from stars as compared to extended objects (which should have emission beyond a radius=2 from center).


"SUPER" SNR: S2SNR, S4SNR, S8SNR, SSNRMAX

From the JHK images, a "super" coadd is generated. The "super" is block averaged, 2 X 2, 4 X 4, and 8 X 8, and boxcar convolved similarly to the individual JHK images. For each LCSB candidate, the SNR is computed for each of these blocked/smoothed images: S2SNR, S4SNR, and S8SNR. The noise corresponds to the background noise from these blocked/smoothed images. Finally, the maximum of the three SNR measures is SSNRMAX. These SNR measures are designed to take advantage of the increased signal-to-noise ratio in the "super" coadd. Low surface brightness objects should have increasing SNR as you go to higher blocked images (depending on the size of the galaxy).


---

Central Surface Brightness

• central surface brightness
  

• central radius=5 surface brightness
  

  white filled circles == galaxies
  red triangles == false detections
  red crosses == artifacts (meteor streaks)
  yellow crosses == unknown objects
  

Blocked JHK-SNR

• Blocked/Smoothed SNR

The dashed blue line represents the threshold that was imposed upon detection: Objects with BSNR < 4 (in all three bands) were considered non-detections. This threshold is imposed to minimize contamination from faint stars. For the objects that survive this cut, the confirmed galaxies, false detections and unknown objects are degenerate in this phase space. Note that most of the objects, real, false and otherwise, are very faint and generally not subject to the level-1 specifications. Another way to view the results is to plot the SNR vs. the uncertainty in the photometry:

• Blocked/Smoothed SNR vs delta_Mag (R=10)
  
• Blocked/Smoothed SNR vs delta_Mag (isophotal)

Indeed most of the objects have at best 20 or 40% photometry (isophotal is for the most part better since the isophotal radius is typically smaller than 10 pixels in radius).

Integrated JHK-SNR

• R=10 Integrated SNR

This space is degenerate for the most part.

"Super Coadd" Blocked SNR

• "Super" SNR with multiple blockings: J band
  

• "Super" SNR with multiple blockings: H band
  
• "Super" SNR with multiple blockings: K band

Most of the confirmed galaxies have SSNRMAX values greater than 7 or 8, while many false detections have values less than this limit. For the individual SSNR values, S2SNR, S4SNR and S8SNR, it is not clear if galaxies separate from stars. SSNRMAX looks the most promising of the SNR parameters to distinguish stars from galaxies.

Disk Scale Length

For the brighter objects, we can compute the disk scale length by fitting an exponential to the mean surface brightness profile. The results are given here:

• Disk Scale Length in arcsec