We propose to modify the MODEX algorithm for estimating direction of arrival (DOA) of plane waves using array of sensors. The MODEX algorithm is a method of direction estimation (MODE) based estimator that conveniently increases the number of MODE estimates, increasing the order of the polynomial, whose roots (candidates) estimate the DOA parameters. To select the better set of candidates, MODEX employs a maximum likelihood (ML) procedure. The proposed method affects the candidate selection process and improves its performance. This performance gain is achieved by replacing the original sample covariance matrix by a new one computed after a noise reduction filtering on the snapshot vectors. The modified selection procedure improves the MODEX performance and reduces significantly the threshold signal-to- noise ratio.