Wavelet-based 1/f-Noise elimination in millimeter-wave radiometer

The output signal of millimeter-wave (MMW) radiometer is usually affected by 1/f-noise that is characterized by self-similarity and non-stationary. Wavelet thresholding techniques are widely used to de-noise. However, soft-thresholding has good de-noising effect but is poor in details preservation, and hard-thresholding is good in details preservation but has poor de-noising effect. To eliminate 1/f-noise from white Gauss noise (WGN), an improved thresholding function is introduced in this paper. This method overcomes the permanent bias in soft-thresholding and the discontinuous point in hard-thresholding. Additionally, a least square estimation (LSE) is introduced to estimate 1/f-type parameters. Simulations and experiments have been carried out to validate this method.