Low roundoff noise structures for 2-D filters

In the implementation of 2-D digital filters using finite wordlengths, errors occur which can have negative effects on the performance of these filters. These effects include changes in the input/output description due to finite accuracy of the filter coefficients, roundoff error due to signal quantization and limit cycles due to overflow and/or quantization. The development of 2-D filter structures which, when implemented using finite wordlengths, are not sensitive to roundoff errors and are free of limit cycles is an active field of research. Some results in this area are briefly reviewed