The covariance methods exert an effect on spatially colored (correlated) noise elimination during direction finding in multiple-input multiple-output (MIMO) radar. However, most of the existing methods seem difficulty to achieve a good balance between accuracy and efficiency. This paper aims at formulating a covariance trilinear decomposition perspective for direction-of-departure (DOD) and direction-of-arrival (DOA) estimation for bistatic MIMO radar. First, the array covariance matrix model is presented for de-noising. Furthermore, the noiseless covariance matrix is rearranged into a trilinear decomposition model. Finally, joint DOD and DOA estimation are linked to trilinear decomposition, which can be easily accomplished by exploiting the existing COMFAC technique. The proposed scheme can exploit the tensor structure of the covariance matrix, and it is attractive from the perspective of computational complexity. Moreover, it can be easily extended to the spatially colored noise scenario. The proposed algorithm is analyzed in terms of identifiability, flexibility, and complexity, and the stochastic Cramer-Rao bound on joint DOD and DOA estimation is derived. The computer simulations verify the effectiveness and improvement of the proposed method.