Abstract: Super-resolution reconstruction with probabilistic motion estimation circumvents the problem that conventional super-resolution methods rely heavily on accurate motion estimation.To improve the effect of its reconstruction,the local structure mode of image is analyzed followed by a definition for local spatial activity given in this paper as a token of the local structure characteristic and a non-local spatial and temporal window function designed to be adaptive in its shape and scale.An adaptive super-resolution reconstruction algorithm of image sequences with structure preserving is further proposed.The experimental results show that the proposed method is better in the reconstruction performance.