The traditional Super-Resolution (SR) algorithms are very sensitive to image registration errors, model errors or noise, which limits their real utility. To enhance the robustness of SR algorithm, this paper improves the traditional SR algorithm from two aspects of image registration and reconstruction. On registration phase, the probabilistic motion field is introduced to prevent the SR algorithm from depending on accuracy of registration. In addition, the Heaviside function is adopted to implement the motion weight mapping, which enhances self-adaption of the algorithm further. On reconstruction phase, a regularized estimation based on Huber norm is used to reconstruct the SR image, which makes the proposed algorithm more stable to minimize the cost function while still robust against large errors. The experimental results show that the proposed algorithm has a good performance on sequence SR reconstruction compared with some existing SR methods.