As we all know, noise always exists in any actual signal processing systems; its strength largely determines the working performance of one signal processing algorithm. Once the noise power is higher, many classical algorithms will lose the ability of the solution. In this paper, we propose a novel robust uncorrelated noise filtering method for direction of arrival (doa) estimation at low signal-to-noise ratio (snr) in coherent environment. Under the assumption of uncorrelated noise the proposed method can remove the uncorrelated noise variances from an array covariance matrix effectively. The characteristics of signal coherency structures and the simulation results using the novel method are presented.