NAME
smooth - to perform Gaussian and other convolutions on image files
SYNOPSIS
smooth file [fw=] [fwr=] [fwc=] [boxcar=] [outfile=] [box=] [file=]
[nodc=]
DESCRIPTION
This routine performs a two dimensional convolution of of the image file with
either a Gaussian, uniform (boxcar), or user supplied function. The full
width at half maximum of the Gaussian can be specified either by specifying
the row and column width (in pixels) separately as fwr= and fwc= or together
using the fw= keyword. If specified, the operation is constrained to a
box. The kernel areas are normalized to one so that the total pixel
count is preserved; i.e. the peak values will
drop.
The `nodc' generates a Gaussian kernel, however, it is biased downward by 0.5
and then the lower levels are normalized so that the net sum of the kernel is
0.
LIMITATIONS
a) The convolution is performed in memory and sufficient resources must be
present for the convolution to be successful. b) The maximum size of the
kernel is 17 pixels, too large a value for fw may exceed this value. c) The
algorithm assumes that the convolution kernel is separable; i.e. that the
functional form of the kernel f(x,y) can be expressed as the product of two
functions g(x)*h(y). d) Convolutions are done in the image domain and can be
slow for large
kernels.
EXAMPLE
smooth alphalyr fw=3.4 outfile=blurry
convolves the image in the file `alphalyr.fts' with a two dimensional
Gaussian with a fwhm of 3.4 pixels and writes the result to the file
`blurry'.
smooth NGC1275 boxcar=4.6
convolves the file NGC1275.fts with a uniform square aperture of 4.6 pixels
on a side and overwrites the input
data.
BUGS
The user specified file option is not ready yet.