The digital image correlation (DIC) method is a non-contact measurement technique based on digital image processing, used to capture the surface deformations of objects. Aiming at the problem of a large computational load in the existing initial guess method, a multi-level grid (MLG) initial guess method based on the transfer correction of calculated subest results was proposed. First, all subsets were divided into grid levels according to the number of subsets and subset intervals, the grid spacing of level 0 was the largest,and the highest-level grid selected the normal subset interval. Then, other initial guess methods were used to perform initial guess and iterative calculation on all subsets of level 0 grid with the largest subset interval. Finally, the initial value of deformation parameter for the higher-level grid subset was guessed based for the deformation parameter for the lower-level grid subset, and the iterative calculation was performed and cycled until the calculation of all levels of subsets were completed. The proposed algorithm is verified by using images from the classic “DIC Challenge”. By comparing the FFTCC integer pixel initial guess method, the scale-invariant feature transform (SIFT) initial guess method based on feature point matching and the reliability-guided initial value transfer method, it is verified that the proposed algorithm effectively improves the overall initial guess accuracy, thereby reducing the number of ICGN iterations and enhancing the overall execution speed of the algorithm.