1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
|
.help craverage Apr98 noao.imred.crutil
.ih
NAME
craverage -- detect CRs and objects using average filter
.ih
SYNOPSIS
\fBCraverage\fR detects cosmic rays and objects using a moving block
average filter with the central pixel plus some number of additional high
pixels excluded and a median of an annulus around the block average box.
It avoids identification of the cores of objects as cosmic rays by
excluding pixels within the detected objects as cosmic ray candidates.
.ih
USAGE
.nf
craverage input output
.fi
.ih
PARAMETERS
.ls input
List of input images in which to detect cosmic rays and objects.
.le
.ls output
List of output images in which cosmic rays are replaced by the block average
value excluding the cosmic ray. If no output image name is given then
no output image will be created.
.le
.ls crmask = ""
List of input and output cosmic ray and object masks. If the mask exists
then the mask values are used to exclude data pixels from the calculations
and zero mask values are candidates for cosmic rays or objects.
Detected cosmic rays and objects are identified in the mask with values
given by the \fIcrval\fR and \fIobjval\fR parameters. If no output cosmic
ray mask is given then no mask will be created.
.le
.ls average = ""
List of output block average filtered images. If no image name is given
then no image will be created.
.le
.ls sigma = ""
List of output sigma images. If no image name is given then no image
will be created.
.le
.ls navg = 5 (minimum of 3)
Square block average filter size given as the number of pixels along an
edge. The value will be rounded up to an odd value to be symmetrical
around the center pixel excluded from the average.
.le
.ls nrej = 0 (minimum of 0)
Number of additional highest pixels to exclude, in addition to the
central pixel, in the block average. The value should be small but it
is needed to deal with cosmic rays that are bigger than a single pixel.
.le
.ls nbkg = 5 (minimum of 1)
Background annulus width around the box average filter in pixels. The
median of the pixels in this annulus is used to estimate the background.
.le
.ls nsig = 25 (minimum of 10)
Square box size for empirical sigma estimates given as the number of
pixels along an edge. The sigma is estimated using percentile points
of the pixels in the box. The size of the box should contain
of order 100 pixels or more.
.le
.ls var0 = 0., var1 = 0., var2 = 0.
Variance coefficients for the variance model. The variance model is
.nf
variance = var0 + var1 * data + var2 * data^2
.fi
where data is the maximum of zero and the average filtered pixel value and
the variance is in data numbers. All the coefficients must be positive or
zero. If they are all zero then empirical data sigmas are estimated by a
percentile method in boxes of size given by \fInsig\fR.
.le
.ls crval = 1
Mask value for detected cosmic rays. It is legal for the value to be
zero to not mark the cosmic rays in the output mask.
.le
.ls lcrsig = 10., hcrsig = 5.
Low and high sigma factors for detecting cosmic rays. These factors
multiply the computed or estimated sigma at each pixel and these threshold
values are compared to the difference between the candidate pixel and the
block average filter value (average of box around the pixel). This only
applies to pixels where the block average filter value is within a
specified threshold of the background estimate; i.e. the average value is
not considered as part of an object.
.le
.ls crgrow = 0.
Cosmic ray growing radius. Pixels detected and marked in the output cosmic
ray mask by the \fIcrval\fR value are increased in size in the mask (but
not replaced in the output image) by also flagging all zero valued mask
pixels within this specified radius with the cosmic ray mask value. This
is done after the detection phase is complete. The separation between
pixels is the distance between pixel centers computed as a real value.
Note a value of at least one is required to affect other mask pixels.
.le
.ls objval = 0
Mask value for detected objects. It is legal for the value to be
zero to not mark the objects in the output mask.
.le
.ls lobjsig = 10., hobjsig = 5.
Low and high sigma factors for detecting objects. These factors multiply
the computed or estimated sigma at each pixel and these threshold values
are compared to the difference between the block average filter value and
the background annulus median. If the values are made very large then
object detection can be eliminated and cosmic rays will be detected
everywhere.
.le
.ls objgrow = 0.
Object detection growing radius. Pixels detected and marked in the output
mask by the \fIobjval\fR value are increased in size in the mask by also
flagging all zero valued mask pixels within this specified radius with the
cosmic ray mask value. This is done after the detection phase is complete
and so object grown pixels are not used in excluding cosmic ray
candidates. The separation between pixels is the distance between pixel
centers computed as a real value. Note a value of at least one is
required to affect other mask pixels.
.le
.ih
DESCRIPTION
\fBCraverage\fR detects cosmic rays and objects using a moving block
average filter with the central pixel and a specified number of additional
highest pixels excluded and a median of an annulus around the block average
box. It avoids identification of the cores of objects as cosmic rays by
excluding pixels within the detected objects as cosmic ray candidates.
The block average filter computes the average of pixels in a box with the
central or target pixel excluded. In addition the \fInrej\fR parameter can
be used to exclude that number of highest remaining pixels as possible
contamination from cosmic rays which are larger than one pixel or possibly
a very nearby additional cosmic ray. The \fInrej\fR value should be kept
small relative to the total number of pixels in the average so that the
average will still be elevated over the median in real underlying objects.
The resulting average is used as the prediction for the value of the target
pixel. The median of the pixels in a square background annulus around the
block average box provides the prediction for the background at the target
pixel.
The target pixel is considered part of an object if the difference between
the average value and the median background exceeds a specified threshold.
If the pixel is NOT considered to be part of an object then if the
difference between the pixel value and the average value exceeds a
different specified threshold it is identified as a cosmic ray.
The thresholds are defined in terms of sigma factors, which may be
different for positive and negative deviations and for object and
cosmic ray identification. The sigma factors multiply an estimate
for the statistical sigma of the target pixel. The estimate is
either based on a noise model or sigma of pixels in a box near the
target pixel.
The \fIcrmask\fR parameter specifies a pixel mask for the image. If the
mask exists then non-zero mask values will be used to exclude pixels from
the average, background median, and empirical sigma estimates. Also any
pixels with non-zero mask values will not be altered either in the output
image or in the final mask. If the mask does not exist then it behaves as
if all mask values are zero. If all pixels in the average box or median
annulus are previously flagged then the estimates will be undefined and
nothing will be done to the output image or mask. Because the task can
use an input mask to mark pixels not to be considered it can be used
in an iterative fashion.
The noise model is given by the formula
.nf
variance = var0 + var1 * data + var2 * data^2
.fi
where data is the maximum of zero and the average estimate for the target
pixel. The coefficients are all given in terms of the data numbers. This
model can be related to common detector parameters. For CCDs var0 is the
readout noise expressed as a variance in data numbers and var1 is the
inverse gain (DN/electrons). The second order coefficient has the
interpretation of flat field introduced variance.
If all the coefficients are zero then an empirical sigma is estimated as
follows. The input image is divided into square blocks of size
\fInsig\fR. The (unmasked) pixel values in a block are sorted and the
pixel values nearest the 15.9 and 84.1 percentiles are selected. These are
the one sigma points in a Gaussian distribution. The sigma estimate is the
difference of these two values divided by two. This algorithm is used to
avoid contamination of the sigma estimate by the bad pixel values. The
block size must be at least 10 pixels in each dimension to provide
sufficient pixels for a good estimate of the percentile points. The sigma
estimate for a pixel is the sigma from the nearest block. A moving box is
not used for reasons of efficiency.
If an output image name is specified then the output image is produced as a
copy of the input image but with the identified cosmic ray pixels replaced
by the average predicted value. Other optional output images are
the average filtered values and the sigma values.
If a mask is specified the detected cosmic rays will be identified with
values given by the \fIcrval\fR parameter and object pixels will be
identified with values given by the \fIobjval\fR parameter. Note that one
does not need to use an output image and the cosmic rays can be replaced by
interpolation in the data using the tasks \fIcrfix\fR, \fIfixpix\fR, or
\fIccdproc\fR.
One final step may be applied to the output mask. The mask values
identified with the \fIcrval\fR and \fIobjval\fR values may be grown
by identifying pixel values within a specified radius with the same
mask value. Note that this step is done at the end and so any pixels
in a preexisting input mask with the same values will also be grown.
Also the grown pixels will not affect the output cosmic ray replaced
image. See \fIcrgrow\fR for a further discussion.
.ih
EXAMPLES
This example illustrates using the \fBcraverage\fR task to
create a mask with cosmic rays and objects identified and displayed.
The image is a CCD image with a readout noise of 5 electrons
and a gain of 3 electrons per data number. This implies variance
model coefficients of
.nf
var0 = (5/3)^2 = 2.78
var1 = 1/3 = 0.34
.fi
.nf
cl> display obj001 1 # Display in first frame
cl> craverage obj001 "" crmask=mask001 var0=2.78 var1=0.34\
>>> crval=1 objval=2
cl> display crobj001 2 overlay=mask001 ocol="1=green,2=red"
.fi
.ih
SEE ALSO
cosmicrays, crnebula, median, crfix, crgrow, crmedian
.endhelp
|
