We propose a variational multigrid method for fast 3D interpretation of image sequences, in which a dense depth map and 3D motion are directly recovered from spatiotemporal change of intensity images without prior matching and estimation. In this paper, we adopt the multigrid methods to efficiently reduce the computational complexity of the variational method, and suggest a new variational formulation to reliably perform the 3D interpretation. We show the efficiency and effectiveness of our method through experimental results with synthetic and real images.