<p>Identifying the number and exact locations of change-points is an important problem in DNA copy number variation detection. The screening and ranking algorithm (SaRa) serves as an efficient change-point detection tool with a computational complexity of <i>O</i>(<i>n</i>). Nevertheless, the performance of the estimated change-points by SaRa depends heavily on the choices of bandwidth and threshold parameters, necessitating meticulous tuning. Furthermore, SaRa’s inherently local nature limits its power and precision in detecting change-points. In this study, we introduce a refined version of SaRa, termed s-SaRa, specifically tailored to enhance the performance of change-point detection. Our approach initially leverages the SaRa algorithm to generate a pool of change-points candidates. Subsequently, a refined algorithm is proposed to improve the initial candidates and a multiple testing procedure is conducted to determine the optimal number of change-points. Simulation analyses underscore the stability (parameter-insensitivity) of s-SaRa, indicating its superior performance in a variety of conditions. Moreover, s-SaRa is found to be more powerful in detecting change-points with weak signals (small jump sizes). We demonstrate the practical efficacy of s-SaRa by applying it to both Coriel and SNP genotyping datasets.</p>