SMOOTH will smooth or convolve the 'source' object with a 2D or 1D Gaussian or boxcar. This is useful for reducing noise, enhancing low surface brightness features, or as the first step for looking for sharp features in an image (by subtracting the smoothed image from the original). Specify the full-width-half-maximum (FWHM) of the Gaussian (boxcar) in pixels with the FW keyword. If desired, you can specify differing widths in the column or row directions with the FWC or FWR keywords. To smooth the image in the column direction only, (each image row is convolved along its extent separately) just specify the width with the FWC keyword. To smooth in the row direction only, (down the columns) just specify the FWR keyword. For Gaussian, the kernel will extend to WID*FWHM, where WID=3 by default, although this can be overridden with the WID= keyword.
To convolve with a box, use the 'BOXCAR' word. Better yet, use the 'RUNMEAN' word, which gives identical results much faster for odd-sized windows, but slightly different results for even-sized smoothing windows. BOXCAR may be removed in future versions, to be supplanted by RUNMEAN.
The convolutions are done in the image domain and may be slow for large widths of the filter. Edges are handled properly. At this time, the FWHMs of the filters are limited to 33 pixel for the Gaussian and 99 pixels for the boxcar.
Since 1-D spectra are treated the same as 2-D images with only one row, use the FWC= keyword when smoothing spectra. Remember that the width of the smoothing kernel (Gaussian or Boxcar) is in units of PIXELS not Angstroms.