In response to the challenges of eliminating the system noises from different viewing directions in a scanning ocular wavefront measurement system, this paper proposes a method to accurately locate the centroids of the spot array of Shack-Hartmann wavefront sensor and reconstruct the wavefront under high noise conditions. At first, the spot array images were globally processed using median filtering and mathematical morphology methods. Secondly, an improved maximum inter-class variance method was employed to determine the threshold for each sub-aperture. The Suzuki contour tracing algorithm was then utilized to identify the spot windows within which the centroids were computed. Finally, the proposed method combining the median area method precisely determined the centroid of the spots. A comparative analysis revealed that traditional methods failed to discriminate real spots under such high-noise conditions, whereas the proposed method could correctly identify them. The wavefront reconstruction errors, evaluated with PV and RMS values were less than 0.015λ and 0.001λ respectively, indicating that the accuracy and stability of the proposed method significantly outperformed traditional approaches.