Skip to main content
U.S. Department of Energy
Office of Scientific and Technical Information

Multiprocessor implementation of two-dimensional digital filters for real time processing

Thesis/Dissertation ·
OSTI ID:6037634

Many practical applications of two-dimensional (2-D) digital filters require high processing speed for a real-time environment. This requirement makes a single processor implementation unfeasible with current technology. A multiprocessor implementation can achieve a high processing rate by dividing the computational load among the processors. Moreover, a multiprocessor implementation can take advantage of the recent developments in very large scale integration (VLSI) technology to achieve high efficiency. The question is how to decompose the computational load among the processors in a VLSI environment to achieve high processing speed and efficiency. This research problem is the major concern of this dissertation. The approach of this research is based on maximizing parallelism and pipelining, minimizing data communications, and having simple and identical computational primitives. In addition, state-space techniques are used as tools in the development of realizations that have natural multiprocessor implementations. This approach insures high processing speed and is amenable to VLSI implementation. In this dissertation, several realization structures for 2-D digital filters are developed and their performance is investigated. First, four new realization structures for the class of separable denominator 2-D filters are developed and their performance is evaluated. The realizations have a cycle-time of one multiplication and one addition, regardless of the order of the filter. Second, two pipelined structures for general FIR and IIR 2-D digital filters are developed. The structures can be mapped onto systolic or wavefront arrays suitable for VLSI implementation. They exhibit high modularity and have regular horizontal and vertical local data communications. Third, a frequency domain implementation of block state-space 2-D digital filters is developed.

Research Organization:
North Carolina State Univ., Raleigh, NC (USA)
OSTI ID:
6037634
Country of Publication:
United States
Language:
English