Impulsive Noise Excision Using Robust Smoothing

Baiqiang Zhang,Junhao Xie,Minglei Sun,Wei Zhou

DOI: https://doi.org/10.1109/lgrs.2021.3093257

IF: 5.343

2022-01-01

IEEE Geoscience and Remote Sensing Letters

Abstract:The impulsive noise originated from regional lightning discharges can dramatically degrade the performance of skywave over-the-horizon radar (OTHR) target detection with generally 10–20-dB increase of noise level in the Doppler domain. As a result, it is necessary to excise impulsive noise in corrupted data while preserving the useful echo signal. Traditional techniques first locate the impulsive noise in slow-time data with a high-pass finite-duration impulse response (FIR) notch filter and then restore the contaminated samples by neighboring good data. However, the effect of impulsive noise is propagated into the subsequent filtered residuals, which will mislead the impulsive noise detection. Instead of the conventional detection-interpolation methods, we propose a novel impulsive noise excision approach based on robust smoothing. The proposed method can robustly smooth the corrupted data and determine the position of impulsive noise with a preset threshold against the residuals of corrupted data subtracting the smoothed signal. The contaminated data at last are replaced by the smoothed signal. Experimental results have demonstrated the effectiveness of our proposed method.