Design and implementation techniques for two-dimensional digital filters. [SECTION performs image filtering with nonrecursive spatial filter in LLLTRAN for CDC 7600]
New design and implementation techniques for two-dimensional (2-D) digital filters are presented. These techniques are characterized by their applicability to a large number of practical problems of current interest in 2-D signal processing. After a brief review of 2-D digital signal processing formalisms, the problem of implementing 2-D nonrecursive filters is investigated. In particular, the sectioning approach to filter implementation is analyzed with emphasis on minimizing the total computation required. An optimal sectioning procedure is formulated, and examples are given which show that sectioning can be significantly superior to other common implementations for a large class of filtering problems. The remainder of the report addresses the problem of designing stable 2-D recursive digital filters. A very efficient design procedure for separable 2-D recursive filters is presented. This procedure is useful in designing quarter-plane recursive filters, and is applicable to a limited subset of design problems. Finally, two techniques are presented for designing stable recursive filters of the more general half-plane variety. These techniques incorporate a recently developed 2-D spectral factorization procedure and a nonlinear optimization to obtain filters to approximate an arbitrary magnitude response, with a zero phase response if desired. Example designs are provided for each technique discussed. 29 figures, 4 tables, 64 references.
- Research Organization:
- California Univ., Livermore (USA). Lawrence Livermore Lab.
- DOE Contract Number:
- W-7405-ENG-48
- OSTI ID:
- 5271145
- Report Number(s):
- UCRL-52370
- Resource Relation:
- Other Information: Thesis
- Country of Publication:
- United States
- Language:
- English
Similar Records
Two-dimensional recursive filter design - a spectral factorization approach
Why filter recursively in two dimensions
Related Subjects
COMPUTER CODES
S CODES
DIGITAL FILTERS
DESIGN
IMAGE PROCESSING
CDC COMPUTERS
DATA PROCESSING
FORTRAN
OPTIMIZATION
SIGNALS
TWO-DIMENSIONAL CALCULATIONS
COMPUTERS
PROCESSING
PROGRAMMING LANGUAGES
990200* - Mathematics & Computers