Abstract: The digital image correlation (DIC) method is a new non-destructive, contactless displacement measurement approach and it can be broadly applied in mechanical engineering with wide application prospect. However, this efficient and convenient displacement measurement method is difficult to apply in fracture mechanics due to the limitation that continuous deformation is required when processing the deformed images with standard digital image correlation methods. Aiming at solving this problem, this paper proposes a novel discontinuous digital image correlation (DDIC) method by introducing the splitted subset model to take the place of continuous model and use it to analyze the discontinuous areas where standard digital image correlation method is not valid in. The displacement of the original pixel points is studied when discontinuities such as cracks occurs, and the crack opening vector is introduced to represent the displacement relationship between the master zone and the slave zone after subset splitting into two parts. Thus the mathematical model of the splitted subset can be established by using the crack opening vector to correlate the master zone and the slave zone, and the corresponding image correlation algorithm can be designed based on this model. Afterwards, the proposed discontinuous digital image correlation method is used to measure the displacements of the deformed images obtained from the cracking process when a tensile test is applied to the notched plate. The research results show that the proposed DDIC method works well in both continuous and discontinuous areas, and when compared with the standard DIC method, the DDIC method is capable to solve the validation problem of image correlation method in discontinuous region, and improves the accuracy for displacement measurement, moreover, the proposed method is able to reconstruct the crack faces and the displacement fields in the vicinity region crack, and the accuracy of displacement measurement can be controlled within the range of subpixel level.