Frequently Asked Questions
Licencing
-
If I get a 100-day licence and then decide to buy an unlimited
licence, do I pay the difference between the price of the latter and
what I already paid?
Yes.
-
I have a licence and have changed machines. Could you issue me
with a new licence key since I will now be using the new machine?
No problem. Just let us know what licence you have, the platform/OS
of your new machine, and its hostid and host name, and we will give you a
new licence key immediately.
[Back to the top]
Deciding
-
Do you support Apple Macintosh, Silicon Graphics...?
Apple Macintosh OS X is now supported.
The following are also supported:
(a) Sun Solaris, Sparc platform, (b) Linux, Intel platform, and
(c) Windows, Intel platform - Windows 95, 98, ME, 2000, NT.
[Back to the top]
-
Do you support a particular wavelet function? Can I do ...
with MR?
Download and have a look through the documentation. Or email us, and
ask. For wavelet functions,
apart from a very wide selection already available, inputing your
own filters is possible too with transform and filtering
programs (using the wvf format of the
Bath Wavelet Warehouse,
which apparently has relocated to
here).
[Back to the top]
-
What does MR offer which is not in Matlab's wavelet toolbox,
IDL's wavelet software, S+Wavelets, public domain packages, ...?
A comparison of what is available in these different packages is
underway. High points of MR include: (a) very sophisticated use of noise
models, (b) extensive availability of redundant wavelet transforms -
best for feature analysis, (c) broad support for, and use of,
multiple scale entropy, (d) unique capability for deconvolution,
filtering based on noise modelling, 3D wavelet transforms, (e) compression
- lossy and lossless grayscale, colour - with memory management for
very large images, and (f) much more besides.
[Back to the top]
-
What image and signal display functionality is available?
The systems for Windows, MR/Materials and MR/Signal, offer image
and signal display functionality.
Otherwise MR programs can be used with other
display environments. Publicly available systems include ImageMagick and
xv. For FITS images, SAOimage is an excellent package, and FView and
others are available for Windows. It is also very easy to use MR
in conjunction with IDL, with the latter serving as display
environment.
[Back to the top]
- FITS extensions - are they handled?
FITS extensions entail multiple frames. The first frame only is
read and processed. Of course breaking the multiple frame FITS file into
single frames allows MR to be used straightforwardly on any one.
[Back to the top]
Advice and Work-Arounds
- Deconvolution, programs im_deconv and mr_deconv:
What format is the PSF to be in?
The PSF is given as an image. This image need not be of large dimensions.
A PSF may be chosen as an isolated point source, e.g. a star. In this
case, extract a subimage around the star and call that the PSF image.
Or if a Gaussian is acceptable, you can
generate it with the -w option in im_simu:
im_simu -f 20.0 -w outnew snpsf out3
This generates a Gaussian of FWHM 20.0 in outnew.fits.
Then run the deconvolution program, maybe limiting the iterations to
50, and checking on progress with the verbose option:
mr_deconv -i 50 -v sn.fits snpsf.fits out
[Back to the top]
- Program mr_decomp.
A problem can arise in image decompression using e.g.
mr_decomp -r 3 IC4678-R-B1.fits.MRC sv3.fits
which partially writes the output file, but in a corrupted manner.
This is due to an Integer to Float conversion bug in the FITSIO
libraries used. Forcing the conversion in the following way avoids
this problem:
mr_decomp -r 3 -t f IC4678-R-B1.fits.MRC sv3.fits
[Back to the top]
- Java Graphical User Interface
There is a known Java bug with the Pentium 4 processor,
which gives the following message: "Java.exe has generated errors and
will be closed by Windows...". From Sun Microsystems:
"Current JDK/JRE 1.1.x users who want to run their JDK 1.1.x based
Java applets or applications on new Pentium 4 systems must
upgrade to JDK/JRE 1.1.8_008 for Windows. Alternatively, Pentium
4 systems are also supported by the Java 2 Platform, Standard
Edition (J2SE) v 1.3.0 and higher."
The version of these Java classes now available on the
software page, and these should be used instead of the
previous class files of the same name when using a Pentium 4.
[Back to the top]
- Program wk_memfilter
Option -U 2 is not yet implemented.
For option -U 1 = fixed alpha, the option -G sets this value.
[Back to the top]
- Program ww_filter
Documentation (p. 311, section 19.5, Examples):
ww_filter -t 18 -T 3 input.fits out.fits
Filtering using the Haar wavelet transform.
This should be:
ww_filter -t 14 -T 3 input.fits out.fits
Filtering using the Mallat biorthogonal wavelet transform.
[Back to the top]
- im3d_convert order of dimensions
In MR/3, the format conversion program im3d_convert
is designed for use with colour images, hence
inputs of dimensions N x M x 3
When the 3rd dimension is greater than 3, a result
is obtained which is not useful.
[Back to the top]
- wk_trans program
In MR/3, the transform program wk_trans
takes the 3rd dimension as the one containing the
sequence of frames.
[Back to the top]
- Short signal
A signal or time series of at least 16 values is needed in MR/Signal.
This is an arbitrary lower limit. Processing a very
small amount of data makes
little sense. Note that signal lengths have no limitation to being
integer powers of 2.
[Back to the top]
Comparisons With Other Systems
-
How does MR differ from the Matlab Wavelet Toolbox?
The Matlab Wavelet Toolbox is strong in its documentation,
and graphical user interface.
Strengths of MR are as follows.
- The à trous wavelet transform algorithm is often
used. This is an excellent compromise between the continuous wavelet
transform (producing a lot of output data when images are
used) and the discrete wavelet transform. It lends itself very
well to what we could characterize as "pattern recognition" type
problems, where features or objects are to to found in the image.
- Another strength of MR is statistical noise modeling. This is
important for imaging based on CCD detectors, where read-out
noise is Gaussian. Faint signal detection has to be based on an
understanding of the image's noise properties. Many noise
models are supported in the MR noise filtering programs.
- All major deconvolution algorithms are supported in MR. For
filtering and for deconvolution, MR's entropy based optimization
and regularization criteria are unique.
- MR's compression programs are based on noise modeling, since we
know that separating a signal well from noise leads to very good
compressibility properties.
- In 1D signal forecasting, an à trous wavelet transform
is an excellent starting point.
- MR lends itself very well to use in other environments. As an
example, entropy and other properties of an image's resolution
scales can be determined, a shell script run to extract appropriate
data, and an input data set prepared for analysis in R (S-Plus). Or IDL
(or PV-Wave)
can be used for image preprocessing and to prepare output results
for reports or presentations.
The above is a selection of important functionality.
[Back to the top]