3D Visual Communications

2.4 3D Scene Representation with Implicit Geometry – Depth-Image-Based Representation

In the previous two sections, we study the geometry-based and image-based representations of 3D scenes. However, we also learn the limitations in each category. Rendering the geometry-based scenes in real time requires expensive hardware and the achieved level of realism is still far below human expectation. Image-based methods encounter the synthesized artifacts problem during the new stereoscopic view generation process, which limits the interactivity. Knowing the pros and cons of each representation, 3D scene representations with implicit geometry (known as depth-image-based representations) are proposed to take advantages from both geometry-based and image-based representations. In other words, it's a hybrid method to include the necessary information from both methods to synthesize new views for 3DTV systems. In this section, we briefly introduce the general concept of depth-image-based representations and provide an overview of depth construction methods which serve as the fundamentals in depth-image-based representations. The details about the construction will be given in Chapter 4 and the compression and encoding and rendering detail will be discussed in Chapter 5.

2.4.1 History of Depth-Image-Based Representation

In early 1990s, all the proposed 3DTV systems were based on the concept of delivering stereoscopic videos, that is, capturing, transmitting, and displaying two separate video ...