We present a new subspace-based direction of arrival (DOA) estimation algorithm for narrow-band sources with high-resolution localization capabilities. The algorithm is based on the invariance property of noise subspace to power of emitters. We specifically compare the resolution of the proposed algorithm with the well-known high-resolution multiple signal classification (MUSIC) method. The proposed method performance is also superior to MUSIC algorithm in the case of correlated sources. Besides, the proposed algorithm is much less sensitive to the power level differences between adjacent sources compared to MUSIC, i.e., very weak sources in vicinity of strong ones can be resolved.