A shift variant filter that automatically locates edge toe and shoulder points was used to reduce the noise in the plateau regions of simulated edges having uncorrelated noise. Edges were generated from Gaussian and triangle spread functions and represent edges produced by linear imaging systems only. Versions of these having flare were also used. For comparison a simple "chop" filter was applied to the same edges. Optical Transfer Functions were generated from the 12 categories of edges, ie. unfiltered, shift variant filtered, and chop filtered for each edge type. Modulation transfer functions of these were compared with MTFs of noiseless perfect edges for each category. The results shows that the shift variant filter improves MTFs best for edges of medium to high noise having flare. When flare does not exist chop filter performance is superior. Surprisingly, no filtering at all results in the best MTF outputs in the case of low noise level edges with flare.
Library of Congress Subject Headings
Imaging systems--Image quality; Image processing--Digital techniques; Optical transfer function
Department, Program, or Center
Chester F. Carlson Center for Imaging Science (COS)
Johnson, David C., "A Shift variant filter applied to edge trace analysis" (1989). Thesis. Rochester Institute of Technology. Accessed from
RIT – Main Campus