Abstract: We developed the pseudo expansion of field-of-view of existing image contents for immersive projection displays like a CAVE or tiled display [Honda et al. 2007]. By using 6 DOF atitude estimation based on optical flow, our approach can reconstruct invisible peripheral images from past image frames with an adaptive depth model. This approach can achieve both real-time processing and reduction of perceptible distortion and discontinuity in the expanded images. It also contributes to using our accessible image contents, like interactive video games, for the immersive displays, and enhancing enjoyment of them.