Transform decomposition in two-dimensional digital filter realization

A novel realization method of two-dimensional digital filters based on well known transforms (Discrete Fourier, Walsh-Hadamard, etc.) is presented. The method proposed here creates realizations which have parallel structure and high throughput rates. It is shown that such realizations are suitable to VLSI implementation. A fixed point error analysis reveals that the realization proposed has better error characteristics than other realizations considered.