Abstract: A common approach for solving source number detection problem is to use an information theoretic criterion, such as Minimum Description Length (MDL).The MDL estimator is known to be nonrobust against temporally or spatially white additive noise assumption. In this paper, a source number detection method used in the presence of unknown spatially nonuniform noise is proposed based on noise reduction of array covariancematrix. The source number estimators using eigenvalue-based method, called the second order statistic of eigenvalues (SORTE), is based on a gap measure of the eigenvalues. The modified covariance matrix is obtained by deducting a diagonal matrix composed of the main diagonal elements of the sample data covariance matrix. The detection method is proposed by the second order statistic of eigenvalues associated with the reduced array covariance matrix. The proposed method is a consistent estimator, robust against deviations from spatially white additive noise assumptions, especially in the low signal-to-noise ratio and small samples. Numerical examples are provided to demonstrate effectiveness of the proposed approach.