Զѧ
ACTA AUTOMATICA SINICA
1997 23 5 Vol.23 No.5 1997



һµ֪ʶָͼ˲
ߡ֡Ԩ
ժҪ˻С任Ķ߶ȱԵʾһͼźűʾ걸ԺӦؽ㷨ȶԣȻС任Ķ߶ȼֵʾĻϣڳ߶ȿռлڶ֪ʶָͼ˲.ʵ˲кǿӦԣȡЧ.
ؼʡС任߶ȱԵʾ߶ȼֵʾָͼ񣬷˲.
A NOVEL KNOWLEDGE BASED NON-LINEAR FILTERING
METHOD FOR FINGERPRINT IMAGES
NI LINBIAN ZHAOQIWANG WENYUAN
(Department of Automation, Tsinghua Univ, Beijing100084)
AbstractThis paper first described the multiscale edge representation of images in the wavelet transform domain, and also discussed the representations completeness and the stability of the corresponding reconstruction algorithm. Then using the multiscale maxima representation based on the wavelet transform, we proposed for fingerprint images a non-linear filtering method which was constrained by several kinds of knowledge in the scale space. The experimental results show that this filtering method has more adaptive performance and can obtain satisfactory results.
Key wordsWavelet transform, multiscale edge representation, multiscale maxima representation, alternate projection algorithm, non-linear filtering
1
S.Mallat1öС任Ķ߶ȼֵ(Multiscale Maxima)ʾ(MMR)ʵͼĶ߶ȱԵʾֱʾάͼf(x,y)ά߶ȿռWf(s,x,y)չõԭʼͼڲͬ߶µݣӶṩĹԭʼͼϢЩϢγɵͼиǿ.Ҫʵֳ߶ȿռеͼȻMMR걸⣬ԼʾĽͶӰؽ㷨ȶ⣬ҪҵMMRԭʼͼֲݵĶӦϵ֪ʶӶʹͼнǿĿԺЧ.ͨS.Mallat1-2ݺZ.Berman3ݵ˵Ĺ˵MMRһͼ񶼾ΨһԣӦؽ㷨ֳܱȶ.ڴ˻ϣָͼ֪ʶڶ֪ʶԼµģڶ߶ȼֵʾָͼ˲.
2С任ͼ߶ȱԵʾ
2.1߶ȱԵʾʵ
ȿһάٽƹ㵽άƽ(x)
