Package: FilteredTransforms Filed on [Indigo]Documentation>FilteredTransformsDoc.tioga Maintainer: Hiller/Crow Date: September 6, 1984 DF file: [Indigo]Top>FilteredTransforms.df Purpose: This package is for doing anti-aliased transformations on AIS format images. Public Interface: FilteredTransforms.mesa If you want to use this package "interactively" (be patient if you try this), here is how: Bringover -p /Indigo/Imager/Top/FilteredTransforms.df FilteredTransforms.load Interp Open FilteredTransformsTest.cm You can now copy commands from FilteredTransformsTest.cm into the interpreter window to create image transformations, or create your own commands based on those in the command file. If you want to use these modules in a program you are writing, see the file TransformTest.mesa for an example of this. One way of creating good anti-aliased images that I was playing with in my last week here was combining this package with the ImageFFT program. If you first filter your file using ImageFFT and then transform it with FilteredTransforms, it seems to produce very good results (better than my filters can alone), though I didn't have time to do any extensive tests. The filter that seemed to work best was the following: FFTFilter yy.ais _ y.ais .3 20 .3 0 BandBoost This creates a trapezoidal low pass filter, symmetrical around the origin. It is intended for use with an image that will be scaled to .3 of its original size. Much to my surprise, an ideal low pass filter did not do a very good job, since it produced a lot of ringing. If you are using this method with more complicated transformations, you should FFT filter it to a frequency that is equal to the largest of m[1][1], m[1][2], m[2][1], or m[2][2] in your transformation matrix (I think). The image looked clearest to me when the "edgeFuzz" was the same width as the cutoff frequency; this may not really be the case so you might want to experiment with it a little. Further information on FilteredTransforms can be found in comments in the other files mentioned here.