A new two-dimensional parallel block adaptive filter with reduced computational complexity

Two-dimensional (2-D) adaptive digital filters (ADFs) for 2-D signal processing have become a fascinating area of the adaptive signal processing. However, conventional 2-D FIR ADFs require a lot of computations. For example, the TDLMS requires 2N² multiplications per pixel. We propose a new 2-D adaptive filter using FFTs. The proposed adaptive filter carries out fast convolution using the overlap-save method, and has a parallel structure. Thus, we can reduce the computational complexity to O(log₂N) per pixel