Abstract: The traditional method uses triangulation and TPS transform reconstruction method, when more than one sampling point, the effect is not good. A 3 dimensional image reconstruction algorithm is proposed based on Quasi Monte Carlo and Taubin smoothing. Quasi Monte Carlo is used for measuring point feature extraction, template construction and calibration are taken, grid model matching and non matching point deformation displacement vector is calculated, the minimum angle principle is taken for carrying on the whole domain vector decomposition of the low-frequency part of the image. Taubin smoothing operator is taken for grid surface updating, match point is obtained in each sampling grid model. The image structure and texture information grid are constructed, triangular mesh surface rendering and 3D image smoothing are realized, 3D reconstruction algorithm is improved. Simulation results show that the algorithm has better 3D reconstruction effect left and right hearts. It has good local conformability with endocardial surface, most of the detail information of the image can be retained, and maximum error of the reconstructed surface is 1.383 4 + 0.939 4mm (P<0.05), meet the demand of image navigation in cardiac surgery.