A novel structure of LWOS filters based on threshold decomposition

In this work, we discuss the restoration of a signal corrupted by additive noise. We aim to develop a smoothing filter which can utilize both the rank- and temporal-order information of the input data. It is well known that the LWOS-filter is one nonlinear filter which can utilize both kinds of information. However, the LWOS filter suffers from rapid growth in the number of parameters as the window size increases. It is impossible to design LWOS filters with a window larger than 5/spl times/5. This window size limitation is a major drawback for image processing. This work presents a novel form of LWOS filter, called the modified LWOS (M-LWOS) filter. M-LWOS filters are defined in the threshold decomposition domain. A M-LWOS filter with arbitrary window size can be easily designed. We present a design method for the M-LWOS filter and show the effectiveness of the proposed filter.