|
HAIT Journal of Science and Engineering B Volume x, Issue x, pp. xxx-xxx © 2005 Holon Institute of Technology | |||||
| |||||
Nonlinear exponential smoothing (NLES) algorithm for noise filtering and edge preservation
Roman E. Goot1, Uri Mahlab2,3, and Reuven Cohen4
1Center for Industrial and Applied Mathematics, | Holon Institute of Technology, 52 Golomb Str., Holon 58102, Israel 2Department of Electrical and Electronics Engineering, Holon Institute of Technology, 52 Golomb Str., Holon 58102, Israel 3Optical Technology and Networking Group, Optical Networks Division, ECI Telecom, 30 Ha-sivim St., Petach Tikva 49517, Israel 4Embedded Software Group, Optical Networks Division, ECI Telecom, 30 Ha-sivim St., Petach Tikva 49517, Israel *Corresponding author: Reuven.Cohen@ecitele.com Received 27 September 2004, revised 16 May 2005, accepted 22 May 2005
| This paper introduces a new smoothing algorithm for time series processing to be implemented in various applications, including, among others, array processing and wireless communication. This new approach has also been tested for noise filtering and edge preserving in images. The algorithm is based on a nonlinear exponential smoothing (NLES) method that shows a significant improvement in the mean squared error and in the step function response, in comparison to other exponential smoothing methods. The interference-elimination results of the proposed algorithm for an input signal containing a noisy image were equivalent to those of moving average filter and Wiener adaptive filter; however, due to its simplicity, efficiency and reduced memory resources needed, the algorithm becomes extremely attractive.
|