In this paper, we introduce a dimension indexing based algorithm SDI for efficient skyline computation. We show that the dominance tests required to determine a skyline tuple can be sufficiently bounded to a subset of current skyline and any skyline tuple can be used to terminate the computation. Our theoretical analysis and experimental evaluation shows that SDI outperforms the state-of-the-art skyline computation algorithms in high-dimensionality domains as well as in low-dimensionality domains.