“Regarding the accuracy issue in the registration process of multispectral images of aviation filter arrays, some scholars have proposed a dual threshold removal method based on matching point position difference surface fitting. This method selects multispectral mid band images as benchmarks, extracts matching points using SIFT algorithm, and constructs a three-dimensional threshold space for position differences by calculating position differences, constructing Delaunay triangulation, and fitting position difference surfaces using IDW algorithm. Finally, precise matching points are selected using this space to complete image registration. Theoretical analysis and experimental results have demonstrated that this method can effectively improve the accuracy of multispectral image registration for aviation filter arrays, providing a new solution for research in related fields.”