A fast and accurate similarity named Hilbert Scanning Distance(HSD)  has recently been presented for point matching. In this study, we improved an efficient algorithm of search strategy for HSD in the large search space. This search strategy is associated with two ideas: a relaxation greedy search, and an accelerating process using Monte Carlo sampling. The experimental results implicate that this improved algorithm is robust and efficient for point matching using HSD. It also makes a tradeoff between accuracy and speed under different requirements.