centerpars
calgorithm = none
The centering algorithm to be employed.
Algorithm "none" is chosen because in most cases accurate centers
are computed by the DAOFIND algorithm and recentering is not necessary.
DAOPHOT users must remember to set calgorithm to "centroid" if they
are using the image cursor as their input coordinate list.
Users of the gauss and ofilter options should realize that their accuracy
depends critically on an accurate estimate of the fwhmpsf.
The centering options are:
none
The initial positions are assumed to be the true centers. The user may
wish to use this option to save time if the DAOFIND task was used to
detect objects.
centroid
The object centers are determined by computing the
intensity weighted means of the marginal profiles in x and y.
This is the recommended centering algorithm for users using PHOT
interactively with the image cursor as input or in the case when the
coordinate list has not already been centered.
gauss (fwhmpsf dependent algorithm)
The object centers are computed by fitting a Gaussian of fixed fwhmpsf to
the marginal profiles in x and y using non-linear least squares techniques.
ofilter (fwhmpsf dependent algorithm)
The object centers are computed using optimal filtering techniques,
a triangular weighting function of half width equal to fwhmpsf and
the marginal distributions in x and y.
cbox = 5.0 (scale units)
The width of the subraster used for object centering in units of the
scale parameter. Cbox needs to be big enough to include sufficient
pixels for centering but not so large as to include a lot of noise.
A reasonable starting value is twice the FWHM of the PSF.
cthreshold = 0 (sigma)
The threshold in sigma above (emission features) or below (absorption)
the data minimum (maximum) in the
centering subraster, below (above) which pixels are not used for centering.
DAOPHOT users should leave this value at 0.0.
minsnratio = 1.0
The minimum signal to noise ratio required for object centering. If
the estimated signal to noise ratio is less than this number the
computed center will be returned with an error estimate.
cmaxiter = 10
The maximum number of iterations performed by the centering algorithm.
All the centering algorithms use this parameter.
maxshift = 1.0 (scale units)
The maximum permissible shift of the center with respect to the initial
coordinates in units of the scale parameter.
If the shift produced by the centering algorithm is
larger than this, the computed center is returned with an error flag.
clean = no
A switch to determine whether or not symmetry-cleaning is to
be performed on the image subraster before centering.
DAOPHOT users should leave this switch set to "no".
rclean = 1.0 (scale units)
The cleaning radius for the symmetry-clean algorithm in units of
the scale parameter.
rclip = 2.0 (scale units)
The clipping radius for the symmetry-clean algorithm in units of
the scale parameter.
kclean = 3.0 (sigma)
The number of standard sky deviations for the symmetry-clean algorithm.
mkcenter = no
Mark the fitted centers on the displayed image.
The centering algorithm parameters control the action of the centering algorithms. The default values have been chosen for efficiency and to give reasonable results in the majority of cases. Note that many of the centering parameters scale with two quantities which are data dependent, the scale of the image, scale, and the standard deviation of the sky pixels sigma.
Using an initial position and scale supplied by the user, a subraster of data cbox / scale pixels wide around the initial position is extracted from the IRAF image. If scale is defined in terms of the half-width half-maximum of the psf then a single value of cbox once chosen will work well for centering objects in images of different seeing and detector characteristics.
The symmetry-clean algorithm may be optionally applied to the image subraster prior to centering. The cleaning algorithm attempts to correct defects in the image subraster by assuming that the image has radial symmetry and comparing pixels on opposite sides of the center of symmetry. The center of symmetry is assumed to be the maximum pixel in the subraster. However if the maximum pixel is more than maxshift / scale from the initial center, the initial center is used as the center of symmetry. Pixels which are inside the cleaning radius are not edited. Pairs of pixels in the cleaning region, r > rclean / scale and r <= rclip / scale and diametrically opposed about the center of symmetry are tested for equality. If the difference between the pixels is greater than kclean * sigma, the larger value is replaced by the smaller. In the cleaning region the sigma is determined by the noise model assumed for the data. Pairs of pixels in the clipping region, r > rclip / scale are tested in the same manner as those in the cleaning region. However the sigma employed is the sigma of the sky background. DAOPHOT users are recommended to leave clean set to "no".
Next the signal to noise ratio of the subraster is estimated. If the SNR < minsnratio, the new center is still computed but an error flag is associated with that object.
New centers are computed using the centering algorithm specified by calgorithm, the data specified by cbox / scale, and points that are cthreshold * datapars.sigma above the local minimum. The recommended centering algorithm is centroid. This centering algorithm computes the marginal distributions in x and y of the extracted subraster. The intensity weighted mean and mean error of the x and y marginal distributions is computed using only points in the marginal array above the minimum data pixel plus a threshold value which is either the mean or datapars.sigma * cthreshold if cthreshold is greater than zero. This algorithm is similar to that used in MPC (Mountain Photometry Code). Note that this is the only centering algorithm for which the choice of datapars.fwhmpsf is not crucial. The centering algorithm gauss computes the new centers by fitting a 1D Gaussian function to the marginal distributions in x and y using a fixed datapars.fwhmpsf. Initial guesses for the fit parameters are made from the data values. The routines then iterate until a best fit solution is achieved. The final choice ofilter employ a variation of the optimal filtering technique in which the profile is simulated by a triangle function of width datapars.fwhmpsf. Detailed descriptions of all the centering algorithms can be found in the Apphot Specifications Document.
If the computed shift in either coordinate < maxshift / scale, the new center is returned but an error flag is set.
1. List the centering parameters.
da> lpar centerpars
2. Edit the centering parameters.
da> centerpars
3. Edit the CENTERPARS parameters from with the PHOT task.
da> epar phot
... edit a few phot parameters
... move to the centerpars parameter and type :e
... edit the centerpars parameters and type :q
... finish editing the phot parameters and type ^Z
4. Save the current CENTERPARS parameter set in a text file ctrnite1.par. This can also be done from inside a higher level task as in the above example.
da> epar centerpars
... type ":w ctrnite1.par" from within epar
BUGS