This paper proposes a two-channel noise estimator for speech enhancement in highly non-stationary environment. The proposed noise estimator utilizes a spatial filter which has a capability of extracting noise information even in speech presence region. We exploit a first-order recursion method with timefrequency varying smoothing coefficients to accurately estimate a noise power spectral density (PSD) in both slowly and rapidly varying regions. The smoothing coefficients are determined by measuring the non-stationarity factor of noise, e.g. degree of noise variation. The non-stationarity factor is derived through a statistical assumption of stationary background noise, which does not need any assumption on the type of non-stationary noise. Since the proposed method efficiently estimates the noise PSD both in stationary and non-stationary regions, the enhanced speech obtained by applying the proposed algorithm to the twochannel enhancement system shows superior performance to conventional approaches in various noise environments.