An efficient design of 2-D FIR digital filters by using singular value decomposition and genetic algorithm with canonical signed digit (CSD) coefficients

SVD has been used to design 2-D FIR filters as several parallel sections of two cascaded I-D filters. This design requires less multiplier than a direct 2- D implementation. Many well established methods can been used to design these 1-D filters. In this paper, we use a combination of SVD and GA to design 2-D FIR filters. 1-D filters coefficients are represented with CSD to improve the throughput rate of the 2-D FIR filters. Another improvement to SVD is used to reduce the number of filter's coefficients.