Construction of Complex Environments from a Set of Depth Images
Enhua Wu, Yanci Zhang, Xuehui Liu · Kluwer Academic Publishers eBooks · 2006
In this chapter, we discuss the research issues and state-of-the-art of construction of complex environments from a set of depth images. A newly proposed automatic construction algorithm is mainly introduced in this chapter. The algorithm proposes a hybrid representation of complex models by a combination of points and polygons. Through the model, a real time walkthrough of a complex scene can be achieved. By the approach, starting from a set of depth reference images, all pixels of the images are classified into two categories, corresponding respectively to the planar and non-planar surfaces in the scene and then all redundant pixels are eliminated according to their sampling rate. For the pixels corresponding to non-planar surfaces, a local reconstruction and resampling process is employed to generate a set of new samples, organized by OBB-tree. For the pixels corresponding to planar surfaces, their corresponding textures are reconstructed and holes in textures are analyzed and filled in pre-processing stage. Under this hybrid representation, a culling algorithm can be employed to greatly improve rendering efficiency and a real time walkthrough of a complex scene can be achieved with no restriction on user’s motion.