Difficulties often arise when digital image correlation (DIC) method is used for serial image matching in nonlinear and large deformation applications. An improved algorithm is proposed by utilizing the concept of fractionized displacement transfer (FDT). For a series of image that records the deformation process,the matching algorithm automatically rechoose an intermediate image as the reference image when the correlation level between the current reference and target images drops below a given threshold. Then the sub-pixel corresponding points in the new reference image are rounded into intergral pixels to facilitate the subsequent DIC matching procedure. A method,which is specifically designed based on the continuum assumption,is proposed for compensating the loss of accuracy caused by the pixel rounding. In addition,a growing method for obtaining initial matching points is proposed,which proved to be able to improve the matching effect and efficiency. The reference images are chosen in terms of the ratio of the number of points in the initial matching set to the number of matched points obtained by the correlation method. Experiment and accuracy analysis demonstrate that the proposed technique is effective and reliable,the matching accuracy is on the order of 0.01 pixel.