A non-unitary joint diagonalization method was proposed to estimate the two-dimension(2D)direction of arrival(DOA)embedded in additive Gaussian noise. Using the rotational invariance property among signal subspace induced by an array of sensors with a displacement invariance structure

a set of spatio-temporal correlation matrices which possess diagonal structure were introduced. By implementing the joint-diagonalization of dimension-reduction correlation matrices

the array response matrix was obtained and the 2D DOA can be achieved. The computational complexity of the proposed iterative algorithm is lower than the alternating columns diagonal centers(ACDC)algorithm and the 2D DOA can be paired automatically. Since each iteration step of the proposed algorithm poses a typical least square problem having a unique closed solution

there is no error accumulation as the ACDC algorithm. Simulations show that the estimation performance of the proposed algorithm is at least higher than the 2D estimation of signal parameters via rotational invariance techniques and ACDC algorithm with 5 dB and 2 dB

respectively.