This paper proposes an efficient lossless video coding scheme based on bi-directional 3D prediction. In this scheme, a video signal at each pel is predicted using not only the current frame but also the motion-compensated previous and following frames. The resulting prediction errors are encoded using context-adaptive arithmetic coding. Coding parameters, such as prediction coefficients and motion vectors, are iteratively optimized for each frame so that a coding rate required for the frame can have a minimum. Experimental results indicate that periodical insertion of B-frames which are encoded using the bi-directional 3D prediction can reduce the overall coding rate by about 3 %.