; IMFIT ;--------------------------------------------------------------- ;! fits gaussians to portions of an image ;# TASK ANALYSIS MODELING ;----------------------------------------------------------------------- ;; Copyright (C) 1995-1999, 2001, 2003-2005, 2008-2011, 2013 ;; Associated Universities, Inc. Washington DC, USA. ;; ;; This program is free software; you can redistribute it and/or ;; modify it under the terms of the GNU General Public License as ;; published by the Free Software Foundation; either version 2 of ;; the License, or (at your option) any later version. ;; ;; This program is distributed in the hope that it will be useful, ;; but WITHOUT ANY WARRANTY; without even the implied warranty of ;; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the ;; GNU General Public License for more details. ;; ;; You should have received a copy of the GNU General Public ;; License along with this program; if not, write to the Free ;; Software Foundation, Inc., 675 Massachusetts Ave, Cambridge, ;; MA 02139, USA. ;; ;; Correspondence concerning AIPS should be addressed as follows: ;; Internet email: aipsmail@nrao.edu. ;; Postal address: AIPS Project Office ;; National Radio Astronomy Observatory ;; 520 Edgemont Road ;; Charlottesville, VA 22903-2475 USA ;----------------------------------------------------------------------- IMFIT LLLLLLLLLLLLUUUUUUUUUUUU CCCCCCCCCCCCCCCCCCCCCCCCCCCCC IMFIT: Task to fit gaussian models to an image INNAME Input image name (name) INCLASS Input image name (class) INSEQ 0.0 9999.0 Input image name (seq. #) INDISK 0.0 9.0 Input image disk drive # BLC 0.0 4096.0 Bottom Left corner of fit TRC 0.0 4096.0 Top Right corner of fit OUTNAME Output image name (name) OUTCLASS Output image name (class) OUTSEQ -1.0 9999.0 Output image name (seq. #) OUTDISK 0.0 9.0 Output image disk drive # NGAUSS 0.0 4.0 Number of components CTYPE 0.0 5.0 Model types; one for each component 0->1=Gaussian 2=Zero level; 3=Zero+slope 4=Zero+slope+curve; 5=see HELP IMFIT Guess of model parameters GMAX Peak of component (JY) 0-> Use maximum value GPOS -50.0 16434.0 (X,Y) position (pixels) 0-> Use position of max GWIDTH -180.0 180.0 (BMAJ, BMIN, PA) of comp. (pixels,pixels,deg) 0->Use clean beam @ Fit model parameters FMAX @ Peak of component (JY) @ 0-> Use maximum value FPOS @ -50.0 16434.0 (X,Y) position (pixels) @ 0-> Use position of max FWIDTH @ -180.0 180.0 (BMAJ, BMIN, PA) of comp. @ (pixels,pixels,deg) @ 0->Use clean beam FSHIFT @ RA/Dec shift to put comp 1 @ on nearest pixel DOMAX $ Solve for GMAX? >0 -> yes $ returns the error DOPOS $ Solve for GPOS? >0 -> yes $ returns the error DOWIDTH $ Solve for GWIDTH? >0-> yes $ returns the error BWSMEAR 0.0 0.1 Bandwidth smearing corr. RADIUS Radius for finding RMS NITER 0.0 4000.0 Maximum # of iteration 0->200*NGAUSS Solve for model parameters? DOPRINT -4.0 132.0 <0 -> Print maps and solutions on line printer >0 -> print on terminal =0 -> maps not printed FITOUT Disk file to save fit info DOOUTPUT -1.0 2.0 >0 -> 1 Catalog residual map OFFSET -1.0 1.0 Cutoff level. 0-> None DOMODEL -1.0 1.0 > 0 => put solutions in a CC file with input image OUTVERS -1.0 MF table version number -1 => none, 0 => new STVERS -1.0 STars output file version -1 => none, 0 => new PBPARM Primary beam parameters: (1) level to believe - <= 0 means do not apply a primary beam (2) > 0 use (3)-(7) ---------------------------------------------------------------- IMFIT Type: Task Use: IMFIT is a task to fit a portion of an image with up to four (4) gaussian components. It can also fit for a baseline term of up to second order. The program estimates the error in the fits 2 ways: using the quality of the fit and using the image rms and theory. The rms for the latter is determined from the image header keyword ACTNOISE (if present and positive) or from fitting the histogram of the full (or partial) image plane. Note that pixels which are exactly zero are not used in this fit, allowing balnked pixels to be REMAGed to zero. The answers are returned in the input guess parameters GMAX, GPOS, GWIDTH and the uncertainties are returned in DOMAX, DOPOS, and DOWIDTH. Adverbs: INNAME......First image name (name). Standard defaults. INCLASS.....First image name (class). Standard defaults. INSEQ.......First image name (seq. #). 0 => highest. INDISK......Disk drive # for the first image. 0 => any. BLC.........Bottom left corner of area of image to fit. TRC.........Top right corner of area of image to fit. Maximum area is 10000 pixels (100x100) OUTNAME.....Residual map name. Standard defaults. OUTCLASS....Residual map class. Standard defaults. OUTSEQ......Residual map seq. #. 0 => highest unique. OUTDISK.....Residual map disk no. 0 => highest with room. NGAUSS......The number of components to use in the fitting. 0->1. Maximum number is four. CTYPE.......Each component type. 0=>1. Two-dimensional elliptical Gaussian 2=Solve for zero level 3=Solve for zero level and slope 4=Solve for zero, slope and curvature 5=Set the six baseline parameters as desired. See EXPLAIN for use of GMAX,GPOS and GWIDTH GMAX........The peak value guess for each component. 0=>Use value with largest absolute value in the BLC,TRC window GPOS........The position (X,Y) guess for components. The values are in pixels in the order (X1,Y1,X2,Y2,X3,Y3,X4,Y4). 0=>Use pixel location with largest absolute value. GWIDTH......The major axis, minor axis and position of major axis guess for components. The values are pixels with degrees for position angle and the order is (MJ1,MN1,PA1,MJ2,MN2,PA2,...etc) 0->Use clean beam, if available; otherwise it will be taken as a circular gaussian of diameter 2. ********* The following are output adverbs: FMAX........The peak value fit for each component is returned. FPOS........The position (X,Y) fit for components is returned. The values are in pixels in the order (X1,Y1,X2,Y2,X3,Y3,X4,Y4). FWIDTH......The major axis, minor axis and position of major axis fit for components are returned. The values are pixels with degrees for position angle and the order is (MJ1,MN1,PA1,MJ2,MN2,PA2,...etc) FSHIFT......The values that one could put in RASHIFT and DECSHIFT to move the first component to the nearest integer pixel. These values include any previous shifts and are corrected for rotation. ********* The following are input/output adverbs: DOMAX.......Flags for GMAX: if > 0 fit this parameter, else hold fixed. Returned value is uncertainty in the fit parameter. DOPOS.......Flags for GPOS: if > 0 fit this parameter, else hold fixed. Returned value is uncertainty in the fit parameter. DOWIDTH.....Flags for GWIDTH: if > 0 fit this parameter, else hold fixed. Returned value is uncertainty in the fit parameter. ********* The following are input adverbs: BWSMEAR.....If > 0, the Clean beam will be smeared by a Gaussian in the radial direction of FWHM proportional to the radius from the pointing position times BWSMEAR. Set it roughly to the channel bandwidth divided by the center frequency. The Clean beam is used as an initial estimate of the source size and is used in the deconvolution attempt to find the true size from the fit size. The peak intensity printed will be corrected for this effect. RADIUS......If = 0, the rms used for error estimates is taken from the image header (keyword ACTNOISE) or found by fitting the full image plane. If RADIUS > 0, then the rms is found by fitting only those pixels within RADIUS of the center of the BLC-TRC box. If RADIUS < 0, then abs(RADIUS) is used as the rms. Pixels which are exactly zero are not used in the fitting for rms. NITER.......The maximum number of iterations to use in the fitting. 0-> NGAUSS * 200. DOPRINT.....<0 -> Plot map, model and residual map and list fit information on the line printer When FITOUT is not blank, DOPRINT=-2 suppresses the page-feed character on page headers and DOPRINT=-3 suppresses page headers and most other header information. DOPRINT=-4 produces output to FITOUT with one line per source component in a special CfA format. =0 -> List fit info in message file only >0 -> Plot map, model, and residual map and list fit information on the terminal FITOUT......Disk file name in which to save the line printer output. ' ' => use scratch and print immediately for interactive jobs - batch jobs use FITOUT = 'PRTFIL:BATCHjjj.nnn' (jjj = job #, nnn = user #). When FITOUT is not blank, multiple outputs are concatenated and the file is not actually printed. DOOUTPUT....>0 -> Catalog residual map with fitted components and write them in a CC file attached to the output image. If DOOUTPUT > 1.5, the components written to the CC file are not deconvolved from the beam. OFFSET......0-> Include all points in fitting area. Otherwise disregard all points less than OFFSET*MAX, where MAX is largest value in fitting window. If MAX is less than zero, disregard all point greater than OFFSET*MAX. DOMODEL.....If true (> 0), put the deconvolved solutions in a new CC file attached to the input image. OUTVERS.....The results are written into an MF (Model Fit) table file with version OUTVERS unless OUTVERS is set < 0. If OUTVERS points at a pre-existing table, the results are appended to the file. OUTVERS = 0 always means to make a new MF table. STVERS......The results may also be written to a STars table attached to the image. -1 => do not do this, 0 => write a new one, > 0 => add to the specified version. PBPARM......Primary beam parameters: (1) Lowest beam value to believe: 0 -> do not do the primary beam correction. This correction is done to the printed parameters only. The beam value used is max (PBPARM(1), that computed from (2)-(7)). (2) > 0 => Use beam parameters from PBPARM(3)-PBPARM(7) Otherwise use default parameters for the VLA (or ATCA where appropriate) (3-7)..For all wavelengths, the beam is described by the function: 1.0 + X*PBPARM(3)/(10**3) + X*X*PBPARM(4)/(10**7) + X*X*X*PBPARM(5)/(10**10) + X*X*X*X*PBPARM(6)/(10**13) X*X*X*X*X*PBPARM(7)/(10**16) where X is (distance from the pointing position in arc minutes times the frequency in GHz)**2. For the VLA, these parms are, by default, given by Perley's fits: 0.0738 GHz -0.897 2.71 -0.242 0.3275 -0.935 3.23 -0.378 1.465 -1.343 6.579 -1.186 4.885 -1.372 6.940 -1.309 8.435 -1.306 6.253 -1.100 14.965 -1.305 6.155 -1.030 22.485 -1.417 7.332 -1.352 43.315 -1.321 6.185 -0.983 For the ATCA, these are by default: 1.5 GHz -1.049 4.238 -0.8473 0.09073 -5.004E-3 2.35 -0.9942 3.932 -0.7772 0.08239 -4.429E-3 5.5 -1.075 4.651 -1.035 0.12274 -6.125E-3 8.6 -0.9778 3.875 -0.8068 0.09414 -5.841E-3 20.5 -0.9579 3.228 -0.3807 0.0 0.0 See EXPLAIN PBCOR for details ---------------------------------------------------------------- IMFIT: Task to fit Gaussian models to an image. DOCUMENTOR: E.B.Fomalont NRAO/VLA RELATED PROGRAMS : SLFIT,IMEAN,MAXFIT,JMFIT,SAD PURPOSE IMFIT fits up to four Gaussian-shaped components to a selected part of an image. One of the components can be a baseline function with a zero level, slope and curvature term. IMFIT is most commonly used to derive the position, peak and integrated intensity and angular size of a source which is not too extended. An initial guess for the parameters, some of which are picked as defaults, must be supplied before running the task. Solution and error estimates are generated and the residual image after the fit can be printed on the line-printer. An arbitrary selection of parameters may be held constant in the solution. The fitting algorithm is based on the subroutine LMDER in the Argonne National Laboratory Subroutine Package and the algorithm uses a linearized least-square solution to obtain the parameters. Occasionally, the solution will converge on an obviously unacceptable fit. If this occurs when fitting one component to the source, try a better first guess. When fitting several components to a blobby source, the fitted parameters may be absurd. Careful selection of fixed parameters will then be necessary. COMMENTS ABOUT SOME PARAMETERS BLC, TRC: The fitting area should be chosen as small as possible; and several disconnected components should be fit separately. The fitting area is limited to an area of 10000 pixels. NGAUSS: The number of components to fit. The maximum number is four and 0-> 1. CTYPE: The component types, placed in a scalar array of length 4. 0->1 Elliptical Gaussian component. 2 Zero level. 3 Zero level and slope. 4 Zero level, slope and curvature. 5 Insert baseline parameters as follows: GMAX = zero level GPOS = slope (intensity per pixel), Orientation (deg N thru E) GMAX = Curvature (intensity per pixel**2), Ellipticity of curvature (-1 to 1), Orientation (deg N thru E) DOMAX, DOPOS and DOMAX are used to hold parameters fixed for Types 1 and 5 only. GMAX: The initial guess of the model intensity may be supplied. The units of GMAX are the same as those in the map. The default of 0 will place the most extreme value in the fitting area (negative or positive) in GMAX for the first component. Any subsequent components with 0 default will be given the value of 0.1 times the extreme value. GPOS: The initial guess of the model position. The location must be given by a pair of pixel coordinates. The default of 0 will insert the location of the extreme value for all components. Note that GPOS has the meaning of slope and orientation for a baseline component. GWIDTH: The defaults for the component widths are generally reasonable; either the clean beam size of a circular beam of two pixels FWHP. Because of poor convergence properties of the algorithm for circular Gaussian models, the task will introduce a slight ellipticity before beginning the fitting. This is not done if either axes is held constant. NITER: The number of iterations, NITER, is defaulted to 200*NGAUSS if it is set to 0. If you are somewhat unsure if your model is reasonable or is converging to an acceptable solution, especially for fits to complicated sources with several gaussians, set NITER=50 and check how the convergence is going. The task has several other termination conditions. If the solution is unchanging to a level of about 0.1%, it will terminate. It some of the fitting parameters are obviously ridiculous, it will also terminate. DOWIDTH: In fitting complicated sources, it is common to hold some of the component diameters fixed in order to obtain reasonable convergence. DOPRINT: Set DOPRINT = -1 in most cases. This produces an automatic hard copy of the solutions and a digital map of the input image, the first guess and the residuals. These maps are most useful for determining the validity of the fit. DOOUTPUT: Catalog the residual map after the fit has been subtracted. If all parameters are held fixed, no fitting is done and a residual map is generated. Only the fitted area is cataloged. A CC file is also written with this output image listing the components. If DOUTPUT > 1.5, the components written are not deconvolved from the beam. If 0 < DOOUTPUT <= 1.5, the components are deconvolved if possible. OFFSET: This adverb permits the exclusion of low valued points when doing the fit. If the extremum value (MAX) in the fitted area is positive, then all points less than OFFSET*MAX are ignored in the fit. If the extremum value in the fitted area is negative, then all points greater than OFFSET*MAX are ignored in the fit. If OFFSET = 0, then all points are used. COMMENTS ON THE USE OF IMFIT SETTING UP THE PARAMETERS For most simple cases the defaults in IMFIT adequately provide starting values. Some examples are as follows: (Always insert the appropriate input map and always set BLC and TRC to the smallest area needed for the fit. The verb TVWINDOW can be used to set the window.) 1. Fit to one Gaussian component. Nothing to specify except flags. 2. Fit to one Gaussian and zero level NGAUSS=2;CTYPE=1,2;GO Note that the zero level information is associated with the second 'gaussian' with CTYPE=2 Additional flags can be specified. 3. Fit to one Gaussian, zero-level and slope NGAUSS=2;CTYPE=1,3;GO Note that the zero level and slope information is associated with the second 'gaussian' with CTYPE=3. Additional flags can be specified. 4. Fit to two Gaussians and a zero level NGAUSS=3;CTYPE=1,1,2;GO Depending on the source complexity it may be important to set some of the fitting flags 5. Fit one Gaussian with a zero level, slope and curvature only in E/W direction NGAUSS=2;CTYPE=1,5;DOMAX=1,1;DOPOS=1,1,1,-1 DOWIDTH=1,1,1,1,-1,-1;GPOS=0,0,0,90; GWIDTH=0,0,0,0,0,90 FLUX DENSITY DETERMINATION: When attempting to obtain the flux density of a well-resolved source, the task IMEAN, which integrates the map values in a specified rectangle, is often more accurate than fitting the source with several Gaussian components and summing the integrated flux densities. PEAK FLUX DENSITY DETERMINATION: The verb MAXFIT, a simple fitting of the peak of a component with a second degree interpolation, is much faster than IMFIT and useful to obtain the approximate peak and position of a component. ERRORS OF PARAMETERS: The error estimates should be regarded as tentative. An estimate of the each error is determined from theory based on the actual rms of the image (neglecting signal portions) or the rms given in header parameter ACTNOISE (if present and positive). AIPS verb ACTNOISE may be used to set this header parameter. See the explain file for SAD for details. Theory gives expressions for the errors in two limiting cases: point source (the beam area > 0.9*the fitted gaussian area) and expanded source (the beam area < 0.1*the fitted gaussian area). The formulae are taken from J. Condon paper 'Errors in Elliptical Gausian Fit', AA, 1996. The intermediate case is handled by interpolation between the two limit cases. It is not clear if this interpolation provides a good estimation of the error in general relation of beam and fitted gausian sizes and position angle. DECONVOLUTION: When fitting to a clean map, IMFIT deconvolves the clean beam from the fitted component size. The nominal deconvolution is obtained by deconvolving the fit from the clean beam. A value of 0.0 means that the source is smaller than the clean beam in some dimension. The minimum and maximum values are obtained by deconvolving the source beam parameters with all 27 combinations of (-1, 0, 1) * uncertainties in the major axis, minor axis, and position angle. The extremas in these parameters over all 27 tries are listed.