ImageFFTDoc.tioga
Tim Diebert, June 10, 1985
Image FFT
CEDAR 6.0 — FOR INTERNAL XEROX USE ONLY
ImageFFT
Plass, Diebert
© Copyright 1985 Xerox Corporation. All rights reserved.
Abstract: Client package for 2 dimensional Fast Fourier Transforms (forword and reverse), intended for image-processing applications. A simple commandTool user interface is also provided.
Created by: Plass, Diebert
Maintained by: Tim Diebert <Diebert.pa>
Keywords: AIS, FFT, filter
XEROX Xerox Corporation
Palo Alto Research Center
3333 Coyote Hill Road
Palo Alto, California 94304
For Internal Xerox Use Only
1.
Executive commands
ClipAIS output ← input <skipLines> <skipDots> <lineCount> <dotCount>
clips a sampled image to the specified dimensions. There are probably more user-friendly ways to do this. "output" and "input" are AIS file names (including extensions).
FFTFilter output ← input <params1> filter1 <params2> filter2 . . .
Applies the specified filters (see below) to the input AIS file to obtain the output AIS file.
FFTShow input <params1> filter1 <params2> filter2 . . .
Displays the log intensity of the FFT of the input AIS file on the color display (ignoring Viewers!), after the specified filters are applied.
Filters:
The following spectral filters may be used with the commands above:
<size> RoundLowPass
a round-shaped low-pass filter
<x> <y> <radius> <m> Spot
scales fuzzy spot around [x, y] by the factor m
<innerRadius> <outerRadius> <edgeFuzz> <m> BandBoost
scales annulus centered on the origin by m
All the filters are applied symmetically in the four quadrants. The values in the frequency domain are in units such that half of the sampling frequency corresponds to a value of 1.0.
Examples:
ClipAIS y.ais ← [pixel]<images>bighumber.ais 150 1060 256 256
FFTFilter yy.ais ← y.ais 0.5 0.5 0.05 0.05 Spot .9 20 .1 0 BandBoost .95 20 .1 0 BandBoost
Limitations:
The FFT program works on square images whose size (height & width) is a power of two. Images that do not conform to this are automatically padded out to the next larger legal size; note that this may result in a lot of wasted space. The largest size tested was 512 square, which takes several minutes. On a machine with four megawords of real memory it might be possible to go to 1024 square, but it would take awhile. It is advisable to experiment with a small patch of the image (say 128 square) to find the appropriate filter before doing the whole thing.