in pipelining. The three stages are performed in 16×16, 8×8 and 4×4, respectively. Smoothness constraints are imposed by using MVs of previous stage as candidates. It must be guaranteed that the candidates for the current stage are well prepared. Therefore each stage is one block away from its previous stage both horizontally and vertically. The organization makes sure that the neighboring blocks used for current stage are processed already in the previous stage so that the three stages can be processed in parallel. For example, to employ MR for current four 8×8 sub-blocks in the block numbered 1, nine neighboring 16×16 blocks have been processed in MP stage, therefore can be used as predictors for MR. The search range for the pipelining is increased to (SRx + (3Nx − 1) • Bx, SRy + (Ny + 1) • By ) due to the parallel multi-access. For inter-frame level, the pipelining of MP, MR and MC is one-frame delayed compared with ME, because the candidates of MP is obtained when ME of the whole frame is complete. ME and the intra-level pipelining in the previous frame compose the inter-level pipelining. We need additional (SRx + (Nx − 1)• Bx)•(SRy + (Ny − 1)• By) search range to perform ME. As a result, Rα is measured using the equation