Abstract: Two-way wave equation based reverse time migration is powerful in imaging complex area. However, its result usually contains imaging noise which poses challenge to seismic interpretation. The imaging amplitude would be distorted by simple noise filter. Wavefield separation imaging condition is an effective way to suppress the imaging noise and enhance the imaging quality. Conventionally, the wavefield separation is achieved in frequency wavenumber domain which cannot be conveniently implemented in time domain reverse time migration. We propose the analytical wavefield extrapolation method which can efficiently separate wavefield into its directional components during the time domain extrapolation. We use this method in reverse time migration and separate both the source and receiver wavefield. And the imaging noise and signal can be separated after applying the imaging condition to the separated wavefield.#br#The wavefield propagation direction is usually defined by temporal and spatial Fourier transform. In the Fourier domain, the wavefield propagation direction is defined by the sign of frequency and spatial wavenumber. In order to efficiently apply the wavefield separation imaging condition in reverse time migration, we extend the analytical time signal into the wavefield and call it analytical time wavefield. Since the analytical time wavefield contains only the positive frequency component, the wavefield propagation direction can be defined by the sign of spatial wavenumber. To avoid the I/O cost in generating the analytical wavefield at every imaging point, we propose an analytical wavefield propagation equation based on the linear relation between the source term and wavefield in wave equation. We solve the proposed equation by finite difference method. Then we separate the source and receiver wavefield into their up and down going components and apply the imaging condition to the separated wavefield. Four imaging components(i.e., imaging component from up going source and down going receiver wavefield, imaging component from down going source and down going receiver wavefield, imaging component from down going source and up going receiver wavefield and imaging component from up going source and up going receiver wavefield)are effectively separated. Then the imaging noise and signal are separated.#br#We use the syncline model and real data to test the proposed method. Numerical example on syncline model shows that the proposed method can effectively separate imaging component from down going source and up going receiver wavefield, imaging component from up going source and down going receiver wavefield, imaging component from down going source and down going receiver wavefield and imaging component from up going source and up going receiver wavefield. When the migration velocity is not accurate, the correlation imaging condition and equivalent wavefield separation imaging condition would generate imaging artifact in central area of the syncline. And the proposed method can eliminate the artifact. Numerical example on real data shows that this method can generate imaging result free from low frequency noise.#br#Wavefield separation imaging condition can effectively separate the imaging noise and signal in reverse time migration and enhance the imaging quality. The proposed analytical wavefield extrapolation method can separate the source and receiver wavefield into their up and down going components. The imaging noise can be suppressed by applying the imaging condition to the separated wavefield. We want to further use this method to generate common angle imaging gather in reverse time migration and provide input data for AVA and migration velocity inversion.