Share Email Print

Proceedings Paper

Selection vector filter framework
Author(s): Rastislav Lukac; Konstantinos N. Plataniotis; Bogdan Smolka; Anastasios N. Venetsanopoulos
Format Member Price Non-Member Price
PDF $14.40 $18.00
cover GOOD NEWS! Your organization subscribes to the SPIE Digital Library. You may be able to download this paper for free. Check Access

Paper Abstract

We provide a unified framework of nonlinear vector techniques outputting the lowest ranked vector. The proposed framework constitutes a generalized filter class for multichannel signal processing. A new class of nonlinear selection filters are based on the robust order-statistic theory and the minimization of the weighted distance function to other input samples. The proposed method can be designed to perform a variety of filtering operations including previously developed filtering techniques such as vector median, basic vector directional filter, directional distance filter, weighted vector median filters and weighted directional filters. A wide range of filtering operations is guaranteed by the filter structure with two independent weight vectors for angular and distance domains of the vector space. In order to adapt the filter parameters to varying signal and noise statistics, we provide also the generalized optimization algorithms taking the advantage of the weighted median filters and the relationship between standard median filter and vector median filter. Thus, we can deal with both statistical and deterministic aspects of the filter design process. It will be shown that the proposed method holds the required properties such as the capability of modelling the underlying system in the application at hand, the robustness with respect to errors in the model of underlying system, the availability of the training procedure and finally, the simplicity of filter representation, analysis, design and implementation. Simulation studies also indicate that the new filters are computationally attractive and have excellent performance in environments corrupted by bit errors and impulsive noise.

Paper Details

Date Published: 9 October 2003
PDF: 10 pages
Proc. SPIE 5146, Optical Metrology for Arts and Multimedia, (9 October 2003); doi: 10.1117/12.499643
Show Author Affiliations
Rastislav Lukac, Slovak Image Processing Ctr. (Slovak Republic)
Konstantinos N. Plataniotis, Univ. of Toronto (Canada)
Bogdan Smolka, Silesian Univ. of Technology (Poland)
Anastasios N. Venetsanopoulos, Univ. of Toronto (Canada)

Published in SPIE Proceedings Vol. 5146:
Optical Metrology for Arts and Multimedia
Renzo Salimbeni, Editor(s)

© SPIE. Terms of Use
Back to Top