1.61-4

	- [BUG] The weighted median algorithm was found buggy. Luckily medians
	were not the default method of estimation, so it is unlikely it caused
	trouble before. Now they are a formidable alternative to maximum-
	likelihood, especially since they have lesser interference with the
	source signals (causing less filtering!). It's worth giving it a try.
	Perhaps in the future these may be made default...

	- [BUG] 'mem' estimator for source model was bungled, often causing
	NaN's. This was traced to occasionbally taking the log of negative
	values. This is now rectified.
	
    	- [NEW] 'imagetool' has added new option "-masks=task1,task2,..."
	Allowing to perform a series of tasks with the currently active
	set of masks. The following tasks are currently implemented:

		forget		Forget all masks.
		add		Add sources as Gaussians with FWHM = radius
		remove		Remove source as Gaussians with FWHM = radius
		pointadd	Add sources as point sources
		pointremove	Remove sources as point sources
		flag		Flag all depfined circular regions
		stack		Stack sources defined in masks.

	Note that the masks have to be defined prior to performing an action
	on them. E.g. 

		'./imagetool -mask=<filename> -masks=<task>'

	Will perform <task> on the mask read from <filename>.
	When sources are added or removed they are done so together with the
	effective beam for the given source size, i.e. together with filtering
	bowls around them if filtering was applied.


	- [DEPRECATE] As a consequence of the above change relating to masks
	'-mask=<filename> -masks=stack' replaces the previous option 
	'-stack=<filename>'.

	- [CHANGE] imagetool's '-mask=' option slightly changed, such that
	it may appear multiple times on the command line, and each time it
	will *add* (not overwrite as before) to the prior masks.

	- [CHANGE] imagetool's option, priorly named as '-clip', is now 
	renamed as '-crop', but the old name is still supported...

	- [FIX] Option '-out=' in 'coadd' made to work correctly.

	- [IMPROVEMENT] Regridding via imagetool's '-regrid' option has been
	somewhat improved with better renormalization, and taking into account
	the small image beam widening as a result of the regridding...

	- [TWEAK] Further small improvements/tweaks to source extraction tool
	'detect'...

	- [IMPROVEMENT] Source extraction now provides more accurate positions
	that are no longer limited to the grid resolution as before. :-)

	- [IMPROVEMENT] Source extraction makes two rounds, inbetween
	re-levelling and noise-rescaling. This allows deeper extraction close
	to the confusion limit. It also makes the routine a lot faster.

	- [IMPROVEMENT] Source extraction actually removes sources as 
	beam-sized Gaussians. This is meant to improve the false finding of
	sources on the wings of bright already identified sources. This happens
	in addition to the flagging arounf a circle with radius of FWHM to
	limit neighbouring sources to a minimal separation of at least one
	FWHM. When writing residuals this is indeed the desired behaviour, 
	since without such flagging the number of counts around zero flux 
	would be artificially enhanced.

	- [NEW] source extractor 'detect' now writes residuals into the same
	location as the original map under a name that is 
	[sourcename].residual.fits. 

	- [SMALL] Source extractions re-levels after each source removeal, not
	only when these exceed the 95% confidence as was the case prior. In
	reality, this change should really have no real impact on the source
	finder...

	- [COSMETIC] Organized help screen for 'imagetool'.

	- [COSMETIC] Added table header to the output of 'detect' (the 
	point-source extraction tool). 

	- [COSMETIC] Checking for update versions improved to include alpha
	beta and release candidate versions also...

1.61-3	

	- [BUG] 'coadd' -name option ended up writing in the default crush
	directory rather than the last component image's directory, which is
	the intended behaviour. 

	- [DEPRECATE] Removed support for old 'RowPolynomials'. This was 
	already uncommented in the configuration for many prior versions, and
	was kept only for legacy purposes. But because it was also buggy, all
	further support is now dropped. One should always use RowBiasDrifts
	instead.

	- [NEW] Added 'Radio' (GLS) projection to the list of projections
	available.

	- [IMPROVEMENT] 'skygradients' wasn't exactly offset-neutral, as it 
	should have been, meaning that in the process of removing sky gradients
	small 'correlated' offsets may have been inserted into the stream.
	Even so things did converge, albeit more slowly than could have. 
	So now, it's all the better

	- [IMPROVEMENT] Speeded up projections by allowing the compiler to
	inline much of the necessary calculation.

	- [IMPROVEMENT] Further speeding up by allowing more code to be
	inlined.

	- Small fixes and improvements to 'imagetool' -deconvolve options. 
	Now it can be used without argument, and deconvolution boost resolution
	2x for all images.

	- Updated 'show' help screen.


1.61-2

	- [NEW] New tool 'difference' added enabling differencing two images.
	When the option '-show' is given, the difference image is displayed
	rather than written as '<sourcename>.diff.fits'.
	Both images can receive additional options via 'imagetool'.
	E.g.

	> ./difference -show -smooth=9.0 first.fits -extFilter=90.0 second.fits
	
	will display the difference between 'firts.fits' (smoothed to 9") and
	'second.fits' (no smoothing by filtering over 90")...
	
	- [MINOR] imagetool's noise clipping is now via '-maxNoise=X' option
	where X specifies the passing dynamic range of noise. E.g. 
	  
	  > ./imagetool -maxNoise=3.0 myimage.fits

	will only keep the part of 'myimage.fits' where the rms noise level 
	does not exceed 3-times the least noisy part of the map. 
	
	- [IMPROVEMENT] Source extraction uses finer regrid (to beam/5), which
	should improve the reliable extraction of *ALL* sources near the
	detection limit, albeit at the price of slower execution...

	- [IMPROVEMENT] Implemented the slower but more accurate regridding 
	routine. The resulting maps are now free from regridding artefacts! :-)

	- [COSMETIC] Reverted to old contrast adjustment method in linear 
	and sqrt scaling modes. This adjustment is available via the keys 'C'
	and 'c' in 'show'. The old-new method turns up or down the contrast
	around the zero level...

	- [COSMETIC] Small corrections (e.g. typos) to console output.

	- [COSMETIC] The printing of angles no longer uses symbols since
	the degree symbol does not show on all terminal types. Thus, reverted
	to more simple dms formatting to be more standard ASCII compliant.

	- [COSMETIC] Updated information screens for all crush tools. These
	used to tell that these tools were for SHARC2 maps, but by now they
	can handle all CRUSH maps (including from minicrush) and others as
	well (e.g. from BoA, or single FITS images).

1.61-1
1.61-b2

	- When masks are defined and displayed, the names are also shown.
	The name labels can be switched on and off using the 'n' key in
	display mode ('show').
	
	- Added capability of stacking sources with 'imagetool' via the 
	'-stack=' option. The argument is the source catalog to be used
	in the same format as mask files (see 'mask.example'). The output
	filename is changed, by default, from '[somename].fits' to 
	'[somename].stacked.fits'.

	- [BUG] Regridding images via imagetool could produced small 
	astrometric shifts. Regridding has been entirely rewritten. The new
	routine results in preserved astrometry and is also faster. Some
	minor regriding artefacts may be produced due to rounding issues.
	The elimination of these would slow down the routine considerably
	and it is therefore not implemented (at this point...).

	- [BUG] 'coadd -offsets' malfunctioned due to an initialization error
	of the amoeba minimizer. This has been corrected and the chi2 minimizer
	has been generally improved.

1.61-b1

	- Switched to nom.tam.fits v0.99-5 Java FITS libraries. This should 
	eliminate a host of FITS problems (e.g. too long value fields for some
	numbers for the liking of GREG even though it is entirely FITS
	standard, and only GREG seems to have trouble supporting it...).

	- [NEW] 'show' display tool can now toggle between linear, logarithmic
	and square-root scaling using the 'x' key. Enjoy! :-)

	- [CHANGE] All image reads now automatically change the working
	directory to the location of the image. Thus, when an output
	name is specified via -name= option, the resulting output will
	be written to the same location as the input image was read from.
	Another effect is that mask files can also be specified relative
	to the path of the image fro which they are applied to. This 
	should add convenience of use...

	- [NEW] Added capability to read single image FITS files (flux
	images). This new feature is still in development, so it may not 
	be very chiseled at the moment. A few warnings to bear in mind:

	   * Presently the reading assumes that the map is in units of
	     Jy/beam. Apply any necessary scaling via -scale= to bring the
	     map inot these units, if you mean to use them...

	   * If BMAJ (specifying a beam major axis in radians) or BEAM
	     (doing the same in arcsecs) is not present in the FITS
	     header, then Nyquist sampling is assumed (i.e. half beam
	     pixelization).

	   * If SMOOTH keyword, specifying smoothing FWHM in arcsec is
	     not present, then unsmoothed images are assumed.

	   * An algorithm tries to guess the noise from the FITS image
	     itself. This algorithm will probably evolve over time...

	- [FIX] Reading and writing files with crush may not have worked
	on Windows platforms because of some hard-coded Unit style path
	specifications. My apologies if this affected you. Unfortunately
	I do (did) very little Windows testing, so if you have comments
	on crush under Windows, please tell me :-).

	- [FIX] Convolution related operations (smoothing, filtering and
	deconvolution) used slightly too small beam patches. This may have
	resulted in images under these operations that were a bit 'edgy'
	but otherwise caused no harm.

