FIR-filter design with spatial and frequency design constraints using evolution strategies

authored by
O. Franzen, H. Blume, H. Schröder
Abstract

A design method for the numerical design and optimization of FIR-filters for video signal processing applications is presented in this paper. The main goal of this filter design process is to achieve a high subjective quality of the filtered image. As the perception of picture quality by the human visual system depends on different filter properties in the spatial domain as well as in the frequency domain, constraints in both domains have to be taken into consideration in the design process, leading to a very complex optimization problem. Furthermore, a low hardware expense for the realization of a video signal filter is also desirable due to the high processing speed that is required for video signal processing applications. We present a set of error functions, each related to one dedicated filter property. The weighted sum of these error functions is a direct measure for the filter quality. Free choice of the weights allows to take direct influence on dedicated filter properties. Evolution strategies, simulated annealing and gradient strategies are compared with reference to their suitability for minimizing the weighted sum of error criteria. The design of both 1-D and 2-D FIR-filters is discussed in this paper. The results of the evolution strategy design technique and standard filter design techniques are compared. For 2-D filter design, results for separable and non-separable 2-D FIR-filters are presented.

External Organisation(s)
TU Dortmund University
Type
Article
Journal
Signal Processing
Volume
68
Pages
295-306
No. of pages
12
ISSN
0165-1684
Publication date
24.08.1998
Publication status
Published
Peer reviewed
Yes
ASJC Scopus subject areas
Control and Systems Engineering, Software, Signal Processing, Computer Vision and Pattern Recognition, Electrical and Electronic Engineering
Electronic version(s)
https://doi.org/10.1016/s0165-1684(98)00079-6 (Access: Closed)