Modeling Based On Multi-Scale Quadtree Image

Abstract

Modeling of a large number of models is a challenging problem in the computer graphics. We propose a novel image-based solution. Our technique introduces quadtree to sample color images and depth images for model or scene. By using quadtree, models have multi-resolution structure. It makes our system have a high rendering speed for lots of models. To show a photo-realistic scene, we use the screen pixel error threshold to select a subset of images that conformed to given conditions and control the accuracy of quadtree subdivision. As the scene is browsed, it interacts well with users. Even if the viewpoints changed arbitrarily, the visual effect is also continuous.