Title :
Training-based optimization of weighted order statistic filters under breakdown criteria
Author :
Huttunen, Heikki ; Koivisto, Pertti
Author_Institution :
Signal Process. Lab., Tampere Univ. of Technol., Finland
Abstract :
A novel optimality criterion in the optimization of weighted order statistic filters is studied. The criterion is based on the concepts of finite sample breakdown point and finite sample breakdown probability which are used as constraints with the more traditional mean absolute error and mean square error criteria
Keywords :
filters; mean square error methods; probability; signal processing; breakdown criteria; finite sample breakdown point; finite sample breakdown probability; mean absolute error; mean square error criteria; optimal parameters; optimality criterion; signal processing; training based approach; training based optimization; weighted order statistic filters; Electric breakdown; Filtering theory; Laboratories; Mean square error methods; Nonlinear filters; Probability; Signal processing; Simulated annealing; Sorting; Statistics;
Conference_Titel :
Image Processing, 1999. ICIP 99. Proceedings. 1999 International Conference on
Conference_Location :
Kobe
Print_ISBN :
0-7803-5467-2
DOI :
10.1109/ICIP.1999.819572
