Arbitrarily variable 2-D filter design and high-speed parallel implementation

Abstract

Frequency transformation-based methods cannot design variable 2-D filters with arbitrarily variable frequency responses. To solve this problem, this paper proposes a linear approach. The new method assumes the variable filter coefficients to be the multi-dimensional (M-D) polynomials of spectral parameters. Then a linear method is proposed for finding the optimal coefficients of the M-D polynomials. Also, the paper shows that the resulting variable 2-D filters can be implemented in parallel form for high-speed signal processing.