3D Video: From Capture to Diffusion

Chapter 12

Coding Methods for Depth Videos

12.1. Introduction

Recent advances in the domain of three-dimensional (3D) cinema and the emergence of new multimedia services such as free viewpoint television (FTV) or 3D television (3DTV) have created a need for new, lighter and less costly 3D video formats. This is a current research focus in both the academic and industrial domains.

The first proposal for 3D video representation was the classic stereoscopic format (stereoscopic 3D (S3D)). This consists of two views of the same scene, recorded by two cameras separated by a certain distance (the baseline). The association of a different view with each of the spectator’s eyes produces a 3D effect by stereoscopy.

Mixed resolution stereo (MRS), followed S3D. This format exploits the theory of binocular suppression, which stipulates that if the two views have a different image quality, the perceived quality of the stereoscopic video will be closer to that of the view with the highest quality [BRU 09]. In practice, this means that if one of the two views is encoded at a lower resolution, the overall quality perceived by the spectator will remain good. Therefore, MRS allows us to reduce the cost (in terms of bit rate) of encoding and transmitting 3D information.

However, two views are not sufficient to deliver wide-range fluid 3D content. This fluidity is necessary for applications such as FTV, where the user has the ability to navigate between different 3D scenes using a controller. Furthermore, ...