Two-dimensional adaptive filtering based on projection algorithm

In this paper, we propose a two-dimensional (2-D) projection algorithm by which filter coefficients can be updated both along the horizontal and the vertical directions on a 2-D plane. Some basic properties of the algorithm are derived. In order to compare with other 2-D adaptive algorithms, it is applied to the identification of 2-D systems with stationary or nonstationary characteristics as well as the reduction of image noise. Results show that the proposed 2-D Modified Projection Algorithm (MPA) achieves good performance for both cases.