Abstract: micro zooming super-resolution imaging is difficult for interpolation and reconstruction, and there are some limitations in the frequency domain model and space domain model based on LSE (least squares estimation). By referring to the mathematical concept of Delaunay triangulation and randomized incremental algorithm, an interpolation algorithm based on Delaunay triangulation in the course of image reconstruction is defined for both real time and accuracy of super-resolution imaging. This algorithm can improve the resolution and reduce the amount of calculation. The results of simulation experiment show that this method is better than the method based on conjugate gradient least square (CGLS) in the speed and error of image reconstruction.