Artifact Identification Algorithms and Parameters

I. Introduction

Bright stars produce various artifacts in the 2MASS images, both inside the scan in which they are detected and in nearby scans. MAPCOR and DB_MAPCOR search for sources created by these artifacts and for real sources that are affected by the artifacts. MAPCOR, running inside the 2MAPPS processing pipeline, works on one band of one scan at a time; thus it only has data from the point sources found in one band of a particular scan. DB_MAPCOR, on the other hand, runs on the entire point source database after pipeline processing is complete; thus it has access to bright stars which are not found in the same scan as their artifacts. MAPCOR searches for sources created by, or affected by, diffraction spikes, persistent latent images, filter and dichroic glints, NICMOS3 "stripes", and halos of bright stars called "parent sources". DB_MAPCOR, using the same artifact identification algorithms and parameters, searches only for sources created by or affected by artifacts that are not limited to the same scan as the parent; namely, diffraction spikes and halos of bright stars. This document describes those algorithms and parameters; another document describes the process used to determine the parameters. I've recently changed the names of some parameters in my web pages to simplify equations and make them more readable. (I've tried to make these names consistent across the pages, but please tell me if I've missed any.) For any who are interested, this page contains a table with the translations between these names used herein and the actual MAPCOR and DB_MAPCOR variable names.

The positional parameters below are all given in units of arcsec (which is the approximate size of the Atlas image pixels); whereas "camera pixels" are approximately two arcsec square. MAPCOR works in band-scan (x,y) coordinates, where (x,y) is the position of the parent source in arcsec relative to the band's zero-point of the scan. The x and y coordinates preserved in the database are in the universal(U)-scan coordinate system, which is in arcsec and independent of band. The y-axis of the U-scan system is parallel to the telescope scan direction, pointing in a northward direction. Converting from each of the three band-scan coordinates to U-scan coordinates involves a 5-parameter linear transformation including x and y scale factors, x and y translations and a rotation angle. All 15 of these parameters change slightly with scan, but the rotation angles remain less than 0.5 degrees regardless of scan. In addition, the relationship between the U-scan system and J2000 RA-Dec axes changes with scan, but the x-axis of the U-scan system never gets more than 3-4 arcmin out of the J2000 equatorial plane. Thus, away from the equatorial poles, to an approximation, x or cross-scan is in the negative RA direction and y or in-scan is in the Dec direction. DB_MAPCOR works in the equatorial coordinate system but rotates the diffraction spikes into the U-scan system of each of the scans in which the possible artifacts are found. (For more information on 2MASS coordinate systems and transformations, please see the POSMAN Subsystem Design Specification.) Note, therefore, that any calculation of MAPCOR or DB_MAPCOR artifact areas using the U-scan or equatorial coordinates found in the database and the parameters herein is only a close approximation to the area calculated by MAPCOR or DB_MAPCOR.

Both MAPCOR and DB_MAPCOR often use the approximate source detection limits as the magnitude limits for artifact searches in each band. If the predicted artifact magnitude is fainter than these limits, within some uncertainty, the program does not examine the possible artifact further. These approximate source detection limits are 17.7 mag in J, 17.2 mag in H, and 16.7 mag in K_s in the northern hemisphere, and 17.7 mag in J, 17.0 mag in H, and 16.6 mag in K_s in the southern hemisphere.

II. Diffraction Spikes

Each band has four diffraction spikes extending in the north, west, south, and east directions (in clockwise order on the images). Since the diffraction spikes vary in length, width, and brightness with hemisphere, band, and direction, each spike in each band and hemisphere has different parameters. The parameters were determined by the patterns of false sources created by the diffraction spikes and faint real sources pushed over the detection threshold by the extra flux, so the artifact search boundaries trace the areas in which the spikes cause false sources, not necessarily the full area of the visible spikes on the images. These spike parameter values are given in Table 1 below.

1. General Parameters

Below certain magnitudes, stars do not cause diffraction spikes bright enough to create false sources. Therefore, MAPCOR and DB_MAPCOR do not search for false or contaminated sources in a diffraction spike if the parent source is fainter than the m_thrD parameter for that spike.

Any sources found within the diffraction spike area are either part of the spike itself or are contaminated by the spike. If a source in the spike area is brighter than m_par + _mD, it is considered contaminated by the spike; if it is fainter, it is considered an artifact of the spike.

2. Spike Length

The calculated length l in arcsec of each spike area, from the center of the parent source, is:

l = l₀ * 10^{[ a *
( m₀´ - m_par ) ]}

where the l₀ and a parameters depend on spike direction, band, and hemisphere, and the m₀´ parameter depends on the source density adjustment (see subsection 3.). In addition, there is a minimum spike length of 2", and a maximum spike length of 4080", or 68', in the north and 3840", or 64', in the south. If the calculated length is less than the minimum, it is set to the minimum value; if it is greater than the maximum, it is set to the maximum value, so no diffraction spike will be longer than 68' in the north or 64' in the south.

3. Source Density Adjustment

The initial value of the m₀ parameter for all diffraction spikes is 6.0. However, as source densities increase, the confusion noise raises the background noise levels such that the diffraction spike lengths decrease. The source density value used by MAPCOR and DB_MAPCOR, src_dens, is simply the number of source detections in the scan for that band, j_n_det, h_n_det, or k_n_det from the scan information table, divided by the survey scan area, 0.85 deg². If the scan's source density is greater than 5883.0 per deg^-2 (?_n_det > 5000), MAPCOR and DB_MAPCOR adjust the spike lengths by shifting the m₀ parameter using:

m₀´ = 6.0 - [ log₁₀(src_dens/5883.0) / d ]

where the d parameter depends on spike direction, band, and hemisphere. The shifted m₀ values result in shorter spikes, as seen in the plots below using parent magnitudes of 3.0, 5.0, and 7.0 as examples.

4. Spike Width

The diffraction spike half-widths now increase with distance from the parent source, making the spikes into wedges that fan out from the center. The equation to calculate the spike width is

w = w₀ + (e * z)

where z is the distance in arcsec from the parent.

5. Diffraction Spike Parameter Table 1 and Plots

	Northern Hemisphere
	J Band							H Band							K_s Band
	m_thrD	_mD	l₀	a	d	w₀	e	m_thrD	_mD	l₀	a	d	w₀	e	m_thrD	_mD	l₀	a	d	w₀	e
N spike	9.75	6.5	91.0	0.127	0.3172	2.0	0.0167	9.25	6.0	76.0	0.132	0.2175	2.0	0.0167	9.00	6.0	62.8	0.139	0.2701	2.0	0.0167
W spike	9.00	6.5	65.4	0.121	0.3517	2.5	0.0278	8.50	6.0	55.0	0.125	0.4641	2.5	0.0300	8.00	6.0	54.2	0.141	0.3223	2.5	0.0300
S spike	9.00	6.5	79.4	0.111	0.3012	2.0	0.0167	8.50	6.0	61.6	0.129	0.4723	2.0	0.0200	8.25	6.0	59.6	0.140	0.5304	2.0	0.0233
E spike	9.75	6.5	86.4	0.114	0.2940	2.0	0.0250	9.75	6.0	65.4	0.126	0.3166	2.0	0.0296	8.75	6.0	75.0	0.112	0.2027	2.5	0.0250
	Southern Hemisphere
	J Band							H Band							K_s Band
	m_thrD	_mD	l₀	a	d	w₀	e	m_thrD	_mD	l₀	a	d	w₀	e	m_thrD	_mD	l₀	a	d	w₀	e
N spike	9.00	6.5	92.2	0.112	0.3620	2.0	0.0200	9.25	5.5	57.0	0.124	0.6244	2.0	0.0167	9.75	6.5	63.4	0.137	0.6943	2.5	0.0167
W spike	9.75	6.5	82.2	0.105	0.3831	2.0	0.0200	9.75	5.5	88.4	0.103	0.3845	2.0	0.0200	8.50	6.5	73.2	0.101	0.3893	2.0	0.0200
S spike	9.75	6.5	85.2	0.122	0.4871	2.0	0.0200	9.75	5.5	60.4	0.127	0.4871	2.0	0.0200	9.00	6.5	56.2	0.137	0.7818	2.0	0.0200
E spike	9.75	6.5	75.0	0.116	0.4258	2.0	0.0200	8.75	5.5	61.2	0.118	0.7503	2.0	0.0200	8.00	6.5	66.2	0.113	0.4923	2.0	0.0200

	Northern Hemisphere
	J band	H band	K band
Spike Lengths
	PostScript version	PostScript version	PostScript version
Spike Widths
	PostScript version	PostScript version	PostScript version
Source Density-Adjusted m₀
	PostScript version	PostScript version	PostScript version
Source Density-Adjusted Spike Lengths for m_par = 3.0
Source Density-Adjusted Spike Lengths for m_par = 5.0
Source Density-Adjusted Spike Lengths for m_par = 7.0
	Southern Hemisphere
Spike Lengths
	PostScript version	PostScript version	PostScript version
Spike Widths
	PostScript version	PostScript version	PostScript version
Source Density-Adjusted m₀
	PostScript version	PostScript version	PostScript version
Source Density-Adjusted Spike Lengths for m_par = 3.0
Source Density-Adjusted Spike Lengths for m_par = 5.0
Source Density-Adjusted Spike Lengths for m_par = 7.0

III. Persistence

Only MAPCOR searches for persistence artifacts, since they are found only in the same scan as the parent source. Even though many of them can be seen in the images following very bright stars, they are not completely point-like and thus not all of them are extracted as point sources. (For more discussions of persistence processing and parameter determination, please see the Persistence Parameters page or the MAPCOR SDS [PostScript file].)

MAPCOR searches for persistence artifacts in J, H, or K_s bands if the parent source is brighter than 11.5 mag, 10.0 mag, or 9.5 mag, respectively. Very bright parents, especially those saturated in Read1, have larger position errors than most 2MASS sources, which in turn cause larger errors in the predicted persistence positions. These persistence sources are also much less point-like than normal sources, because the saturated core of the parent is so large, and thus the persistence detections are also more spread out. Therefore, the persistence search radii and predicted persistence position errors for parents brighter than 5.0, 4.5, or 4.0 mag in J, H, or K_s, respectively, are larger than those for fainter parents.

1. Algorithms

The predicted position of a persistence source found at the location of the i^th frame offset following the parent source is approximately:

x_iP = x_par + ( i * xoffs)
y_iP = y_par + ( i * yoffs)

where (x_par,y_par) is the position of the parent, xoffs is the scan's average frame-to-frame cross-scan offset, and yoffs is the scan's average frame-to-frame in-scan offset. The signs of the offsets depend on the scan direction. These offsets are read from a POSFRM output file for each scan; typical values are, in arcsec, for north-going scans:

xoffs ~ 0.5"
yoffs ~ 83"

and for south-going scans:

xoffs ~ -0.5"
yoffs ~ -83"

The predicted magnitude of the i^th persistence source is approximately given by:

m_iP = m_par + _miP

where the

_miP values are found in Table 2 below. Note that

_miP is not well-determined for values of i > 7 because there are few of these sources and their parents are very bright, so the persistence sources are not point-like and their magnitudes can have large errors.

Using Numerical Recipes routines, MAPCOR determines the persistence probability for each source within 6" of the predicted persistence positions, or within 14" if the parent is brighter than 5.0, 4.5, or 4.0 mag in the J, H, or K_s bands, respectively. The probability that a certain source is a persistence artifact of the parent source is determined from a ² quantity, computed as:

² = [(x_s - x_iP)² / _x² ] + [(y_s - y_iP)² / _y² ] + [(m_s - m_iP)² / _m² ]

where x_s, y_s, and m_s are the possible persistence source's position and magnitude, and x_iP, y_iP, and m_iP are the predicted persistence position and magnitude. The quantities _x², _y², and

_m² are the sums of the squares of the possible persistence source's uncertainties in x, y, and magnitude (found in the source data), the parent source's uncertainties in x, y, and magnitude (also found in the source data), and the uncertainties in the empirically predicted positions and magnitudes, which are 4.0" in x and y, or 10.0" for very bright parents, and 1.0 mag.

MAPCOR keeps searching for persistence with values of i up to the point at which m_iP, within its uncertainty, is fainter than the approximate source detection limits of the survey. Since the predicted magnitude uncertainty is approximately 1.0 mag, the search is ended when m_iP > 18.7 mag in J, 18.2 mag in H, or 17.7 mag in K_s in the northern hemisphere, or 18.7 mag in J, 18.0 mag in H, or 17.6 mag in K_s in the southern hemisphere.

2. Persistence Parameter Table 2

	Northern Hemisphere
i	1	2	3	4	5	6	7	8	9	10
J band _miP	5.8	7.6	8.4	8.8	9.1	9.4	9.8	10.5	12.0	13.0
H band _miP	6.3	7.8	8.4	8.8	9.1	9.2	9.4	11.0	12.0	13.0
K_s band _miP	6.3	7.8	8.4	8.8	9.1	9.4	9.6	10.5	12.0	13.0
	Southern Hemisphere
i	1	2	3	4	5	6	7	8	9	10
J band _miP	6.2	8.1	8.9	9.5	10.0	10.4	10.8	11.5	12.0	13.0
H band _miP	5.8	7.1	7.6	8.3	8.6	8.9	9.3	10.2	12.0	13.0
K_s band _miP	6.3	8.0	8.8	9.3	9.4	9.5	9.8	11.0	12.0	13.0

IV. Filter and Dichroic Glints

All filter and dichroic glints are located close to the parent source, so they will be found only in the same scan as the parent and DB_MAPCOR does not search for them. There are 2, 1, and 2 glints in the J, H, and K_s bands in the north and 4, 2, and 2 glints in the south, respectively. The brightest glints are normally closest to the parent source, and for very bright parents, all glints tend to be lost within the halo of the parent.

1. Algorithms

MAPCOR marks a source as a glint if it is found within r_gl arcsec of the glint's expected position (x_gl, y_gl) and has a magnitude within ± _mgl mag of m_par + _mG, where m_par is the magnitude of the parent and r_gl, _mgl, and _mG depend on the glint number, band, and hemisphere. The expected position is:

x_gl = x_par+f_gl
y_gl = y_par+h_gl

where (x_par, y_par) is the position of the parent, and the f_gl and h_gl parameters also depend on the glint number, band, and hemisphere.

2. Glint Parameter Table 3

	Northern Hemisphere
	J Band		H Band	K_s Band
	glint #1	glint #2	glint #1	glint #1	glint #2
f	1.5	3.0	1.5	1.5	-4.0
h	-14.0	-28.5	-14.0	-14.0	16.5
r_G	2.0	3.0	2.0	2.5	2.5
_mG	6.7	9.0	7.9	7.8	7.2
_mG	0.9	2.0	1.6	1.8	1.3

	Southern Hemisphere
	J Band				H Band		K_s Band
	glint #1	glint #2	glint #3	glint #4	glint #1	glint #2	glint #1	glint #2
f	-2.5	1.0	-6.5	-10.0	-2.5	1.0	-2.5	1.0
h	11.0	-13.5	22.0	33.5	11.5	-13.5	11.0	-13.5
r_G	3.0	1.5	3.5	5.0	3.0	1.5	3.0	2.0
_mG	6.5	9.0	9.0	12.0	8.0	8.2	7.3	7.5
_mG	1.0	2.0	2.5	3.0	2.0	2.0	2.4	1.5

V. NICMOS3 Stripes

As with persistence artifacts, the NICMOS3 horizontal stripes are found only in the same scan as the parent source, so DB_MAPCOR does not search for them. Only sources brighter than the approximate source detection limits in each band minus 11 mag can create stripes that are visible on the images, so parents fainter than 6.7 mag in J, 6.2 mag in H, and 5.7 mag in K_s in the north or 6.7 mag in J, 6.0 mag in H, and 5.6 mag in K_s in the south skip this step in the artifact processing. The three stripes per parent are found at the parent's y position and at y ± 256", and extend horizontally across the entire image. They are 8" wide (full width, north to south), and often have small ghosts at the locations (x ± 0, 256"), (y ± 0, 256"). Anything found within the stripe area is flagged as contaminated by the stripe.

VI. Confusion

MAPCOR and DB_MAPCOR search for two different types of confusion, artifact confusion and photometric confusion. Artifact confusion refers to false source extractions created by the bright parent star, which cluster close to the core of the parent's image. On the other hand, photometric confusion refers to real sources located near a bright parent star which affects their photometry.

The artifact confusion radii is scaled with the magnitude of the parent, but the photometric confusion radii depends on both the magnitudes of the parent and the fainter neighboring source and thus is calculated separately for each possible confusion pair. The artifact confusion radius is usually the smaller of the two; any source inside this radius is considered an artifact. Any source outside the artifact confusion radius but inside the photometric radius is considered a real source with photometry affected by the parent.

1. Artifact Confusion

a. Algorithms

Analysis has shown that only saturated parent sources tend to create confusion artifacts, false source extractions usually clustered in a ring just outside the saturated core of the parent star image. Therefore, MAPCOR and DB_MAPCOR search for artifact confusion in Read1 and/or Read2-Read1 only when the parent is saturated in that read type.

However, there often are no good Read2-Read1 data for sources saturated in Read2-Read1, and thus no explicit indication that the parent was saturated in Read2-Read1. To deal with this situation, the programs also use magnitude thresholds m_sat to determine if the parent should have been saturated in Read2-Read1. These thresholds are found in Table 4 below. If the parent is brighter than the threshold in that band, the program will search for artifact confusion in that band.

Unfortunately, these same magnitude threshold values were used to determine if the parent should have been saturated in Read1 as well, instead of these values minus ~ 3 mag. The artifact confusion parameters for Read1 parents, however, make the artifact confusion radii very small for parents with magnitudes between the correct thresholds and those actually used: 4.0" at 5.0 mag in J, 3.6" at 4.7 mag in H, and 4.5" at 4.1 mag in K_s in the northern hemisphere, and 4.6" at 5.0 mag in J, 3.6" at 4.6 mag in H, and 3.8" at 4.1 mag in K_s in the southern hemisphere. Most of the Read1 sources found inside these radii around the parent are probably caused by missed Read1 frame merges.

The equation to calculate the artifact confusion radius for a read type (Read1 or Read2-Read1) is

r_iAC = r_0AC * 10^{[
b * ( m_i0AC - m_par ) ]}

where the r_0AC and b parameters depend on hemisphere and band, and the m_i0AC parameter depends on hemisphere, band, and read type i. In addition, there is a minimum radius of 1", and a maximum radius of 4000", or 66.67', in the north and 5000", or 83.33', in the south. If the calculated radius is less than the minimum, it is set to the minimum value; if it is greater than the maximum, it is set to the maximum value.

b. Artifact Confusion Parameter Table 4 and Plots

	Northern Hemisphere
	J Band	H Band	K_s Band
m_sat	8.0	7.7	7.1
r_0AC	12.90	10.20	8.50
b	0.202	0.203	0.171
Read1 m_0AC	2.5	2.5	2.5
Read2-Read1 m_0AC	6.0	6.0	6.0
	Southern Hemisphere
	J Band	H Band	K_s Band
m_sat	8.0	7.6	7.1
r_0AC	15.06	9.46	7.74
b	0.207	0.201	0.194
Read1 m_0AC	2.5	2.5	2.5
Read2-Read1 m_0AC	6.0	6.0	6.0

Northern Hemisphere	Southern Hemisphere

PostScript version	PostScript version

2. Photometric Confusion

a. Algorithms

Any source located near another source may have magnitudes that are affected by the neighbor's light; when the photometry is affected at the 5% level or more, the source is flagged as having photometric confusion. Parent sources in this type of confusion can have magnitudes as faint as the approximate source detection limits, and MAPCOR and DB_MAPCOR examine all fainter nearby sources for photometric confusion.

The photometric confusion radii depends on both the magnitudes of the parent and the fainter neighboring source and thus is calculated separately for each possible confusion pair. The equation to calculate the radius for a pair of sources is

r_PC = r_0PC * 10^{[
c * ( m_fs - m_par ) ]}

where m_fs is the magnitude of the fainter source, and the r_0PC and c parameters depend only on band. There is a minimum radius of 1", and a maximum radius of 1200", or 20.0'. If the calculated radius is less than the minimum, it is set to the minimum value; if it is greater than the maximum, it is set to the maximum value. Additionally, all sources with a blend flag greater than zero are marked as having photometric confusion in that band, since they have been passively deblended.

b. Photometric Confusion Parameter Table 5 and Plots

	J Band	H Band	K_s Band
r_0PC	4.678	4.380	3.980
c	0.1165	0.1148	0.1243

The plots below show the maximum photometric confusion radius for a certain parent magnitude in each band and hemisphere, where the second fainter source has a magnitude equal to the approximate source detection limits in that band and hemisphere.

Northern Hemisphere Southern Hemisphere

PostScript version PostScript version

T. Evans - IPAC
Last Update - 3 Feb 2003