Summary: | 碩士 === 國立中正大學 === 資訊工程研究所 === 82 === Stack filters are nonlinear filters which are based on positive
Boolean functions as their window operators. Each positive
Boolean function gives distinct effect on the image. First, we
define two new filters that are called (m, n) rank order
filters and (m, n) fuzzy rank order filters where m<n. We set a
counter to count the number of 1s for any input signal of
window width 2N+1. If the counter has more than n, then the
output is 1. If the counter has less than m, then the output is
0.The difference of (m, n) rank order filters and (m, n) fuzzy
rank order filters is at the value of the output when the
counter is between m and n. For (m, n) rank order filters, we
still take 0 or 1 as the output when the counter is between m
and n. But we take a continue value between 0 and 1 as the
output for (m, n) fuzzy rank order filters when the counter is
between m and n. In this algorithm, We propose the cube
membership function. And we adjust the shape of the cube
membership function by neural learning. From those, we can find
an optimal (m, n) fuzzy rank order filter. Finally, we also can
get the optimal stack filter.
|