Virtual Viewpoint Rendering Based On Image Grid

In the virtual view rendering technology based on image grid,the number of initial grids may exceed the range that the image hardware can tolerate.Therefore,it is necessary to study grid simplification technology to deal with the problem of slow speed of virtual models generated by complex models.Most of the simplification techniques are mainly based on topographic research.They do not take into account the discontinuity of depth maps;they do not handle high-resolution depth maps very well.This thesis aims to solve these problems.In the traditional mesh generation algorithm based on balanced quadtree,although the standard unit can solve the non-equilibrium quadtree problem,the complexity of the judging method with neighboring square series no more than one level is too high.This thesis has improved the standard unit model and used the position and quantity information of the dangling node(common node)of the square to judge which standard cell style to use.This kind of judgement can speed up the grid generation and reduce the time spent on special cases.It can not only guarantee the principle of balanced quadtree,but also can deal with the problem of transition between square grids.In this thesis,the depth of the depth image is added to the five mesh simplification models to be applied to the discontinuity of the depth image.Five depth-based mesh generation methods are depth map extraction pixels,progressive meshing,quadratic error metric meshing,block-based quadtree decomposition,and meshing based on standard cells and quadtrees.Generation algorithm to achieve virtual view rendering based on image grid.The results show that the improved virtual view generated by the grid algorithm based on standard cells combined with quadtree grid has higher efficiency and better quality.