A modified approach based on covariance differencing technique for a uniform linear array is proposed to the problem of direction finding of narrowband signals under color noise background. By assuming a Hermitian symmetric Toeplitz matrix for the unknown noise, the array covariance matrix is transformed in an appropriate way and the noise component is eliminated finally. It is shown that the presented transform-based covariance differencing algorithm works well when the signals are uncorrelated or just two signals are coherent. When the number of coherent signals is larger than two, the proposed scheme combined with spatial smoothing technique is further investigated. New approach provides accurate direction of arrival estimation and avoids generating "phantom" angles relative to the standard method which requires further procedure to determine actual directions. Therefore the proposed approach operates at a lower computational complexity compared to the standard method. Simulation results demonstrate the effectiveness of presented algorithm.