Abstract: Traditional mesh simplification and multiresolution methods cannot work on large models whose data sizes exceed the main memory.In this paper,we present a new out-of-core simplification and level-of-detail （LOD） control method based on an idea of mesh cutting.In our method,a large model is cut into small patches and these patches are simplified separately in one single pass,by utilizing edge correspondences. Furthermore,view-dependent refinement of the whole model is also supported by a multiresolution structure—vsplit forest.When the model is rendered,the local details are rapidly extracted from the structure,and the LOD of the whole model is distributed continuously according to gaze-points of the viewers.