1.1.1.0 image.fitsky - Function

nrao

nrao

NRAO Home > CASA > CasaDocument

[next] [prev] [prev-tail] [tail] [up]

image.fitsky - Function

1.1.1 Fit 2-dimensional models to the sky
Description

This function fits 2-dimensional models to regions of the sky. You can fit multiple models simulataneously, although only Gaussian models are currently implemented. The fit is returned as a record that can be loaded into a Componentlist tool with the fromrecord() function.

If fit=T the fit is done and returned. If fit=F the fit is not done. Instead, the initial auto-determined estimate is returned (this is used as the starting guess of the actual fit); in this case, arguments pixels, pixelmask, converged, estimate and fixed are ignored. The auto-fit can only be determined for a single model.
The pixels array is returned to you (in the return record) with the residual of the fit. The pixelmask array is returned to you (also in the return record) with the pixel mask of the specified region (this reflects both the underlying pixel mask that you applied [if any] as well as the shape of the region).
If the fit converged, converged=T in the return record.
The region-of-interest must be 2-dimensional and it must hold a region of the sky. Any degenerate trailing dimensions in the region are discarded.
The models argument specifies the simultaneous model types you would like to fit to the specified region. Currently you can only choose from ‘gaussian’ (minimum match). Specify the models argument as a vector of strings. Thus, to fit 3 simultaneous gaussians, you would give models=[’gaussian’,’gaussian’,’gaussian’].
An initial automatically determined estimate of the parameters of the model can be made for you (only for single model fits). However, if this turns out to be poor, or if you want to hold some parameters fixed, then you can enter your estimate with the estimate argument to which you supply a record associated with a Componentlist tool (the same type of record that is returned by this function). Note that an integrated flux is specified in the estimate of a Gaussian component. The estimate doesn’t have to be the same type as the model you are fitting. For example you can give a Point model estimate for a Gaussian fit. The other parameter estimates will be filled in for you. The function findsources can be used to get rough point source estimates. You must supply one estimate component for every model component that you wish to fit when fitting multiple simultaneous models.
When you fit a model, you can choose which parameters you hold fixed in the fit and which you solve for. The fixed argument enables you to specify which parameters of the fit you hold fixed. It is entered as a vector of strings, one string for each model specifying which parameters of the fit are fixed.
For gaussian models, that string should contain letters chosen from
- ‘f’ - hold the peak flux (intensity) of the model fixed
- ‘x’ - hold the X location of the model fixed
- ‘y’ - hold the Y location of the model fixed
- ‘a’ - hold the major axis width fixed
- ‘b’ - hold the minor axis width fixed
- ‘p’ - hold the position angle fixed
If you hold any parameters fixed, you may need to specify an input estimate. As currently implemented, the effect of holding constant the flux of a Gaussian model is confusing. The input estimate is specified as an integral flux. This is used to determine the peak flux of the Gaussian model. It is this peak flux which is kept constant. The returned model’s integral flux may be different than the specified integral flux because other parameters (such as the shape) of the Gaussian model may have changed.
Often when fitting, it is advisable to exclude noise pixels. The includepix (pixel value range to include) or excludepix (pixel value range to exclude) arguments (you can’t specify them both) allow you to select a subset of the pixels for the fit. If you only give one value for either of these vectors, say includepix=[b], then this is interpreted as includepix=[-abs(b),abs(b)].
For total intensity (Stokes I) images, if neither of these are specified and the peak in the region is positive, then all pixels below zero are discarded. If the peak is negative (a negative source), all pixels above zero are discarded. For other Stokes parameters all pixels are included by default.

Return value is a record with 4 elements. The keys are ’return’, ’pixelmask’, ’pixels’ and ’converged’.

’return’ is a componentlist record which can be used with the componentlist tool. (The returned flux values are integral, not peak values.)
’pixelmask’ is a boolean array that defines on the image where the residual values are invalid
’pixels’ is a float array containing the residuals after subtracting the fitted components
’converged’ is a boolean value representing the convergence of the fit.

Arguments


Inputs
region		The 2-D region of interest to fit. Default is whole image.
		allowed:	any
		Default:	record
mask		OTF mask, Boolean LEL expression or mask region. Default is none.
		allowed:	any
		Default:	variant
models		The model types to fit simultaneously. Default is [’gaussian’]
		allowed:	stringArray
		Default:	gaussian
estimate		Initial estimate for the model (from componentlist.torecord()). Default is to use an internal estimate.
		allowed:	any
		Default:	record
fixedparams		Parameters to hold fixed per model, String or vector of strings, one per model. Choose from “fxyabp”. Default is all parameters vary.
		allowed:	stringArray
		Default:
includepix		Range of pixel values to include. For default behaviour see discussion.
		allowed:	doubleArray
		Default:	-1
excludepix		Range of pixel values to exclude. For default behaviour see discussion.
		allowed:	doubleArray
		Default:	-1
fit		Do fit or get estimate
		allowed:	bool
		Default:	true
deconvolve		Deconvolve from restoring beam
		allowed:	bool
		Default:	false
list		List informative messages to the logger
		allowed:	bool
		Default:	true

Returns
anyrecord

Example

"""
#
print "\t----\t fitsky Ex 1 \t----"
ia.maketestimage(’image.large’, overwrite=true)
innerquarter=ia.setboxregion([0.25,0.25],[0.75,0.75],frac=true)
ia.fromimage(outfile=’image.small’, infile=’image.large’,
             region=innerquarter, overwrite=true)
ia.close()
ia.open(’image.small’)
cl1 = ia.fitsky(models=["gaussian"])
print cl1.keys()
# ’pixels’,’pixelmask’,’converged’,’return’
print cl1[’return’] # ’return’ field holds a componentlist record
if cl1[’converged’]:
  ia.close()
  ia.open(’image.large’)
  cl2 = ia.fitsky(models=["gaussian"], estimate=cl1[’return’])
  print cl2[’return’]
ia.close()
#
"""

In this example we assume the image is 2-dimensional. We fit the first image using the internal starting estimate for the model parameters. We fit the second image using the result of the first fit as the starting estimate.

[next] [prev] [prev-tail] [front] [up]

Please send any comments or questions about CASA or AIPS++ to aips2-requests@nrao.edu

Copyright © 2007 Associated Universities Inc., Washington, D.C.

This code is available under the terms of the GNU General Public Lincense

Home | Contact Us | Directories | Site Map | Help | Privacy Policy | Search

Updated daily during alpha development.