Chapter 4
Multi-View Depth Image Based Rendering

“I make use of painting to render thoughts visible.”


René Magritte, Belgian surrealist painter.

This chapter focuses on the rendering of synthetic images using multiple views. First, two different rendering techniques are investigated: a 3D image warping and a mesh-based rendering technique. Each of these methods has its limitations and suffers from either high computational complexity or low image rendering quality. Afterwards, we present two novel image-based rendering algorithms addressing the aforementioned limitations. The first of these two algorithms is an alternative formulation of the relief texture algorithm which was adapted to the geometry of multiple views. The proposed technique avoids holes in the synthetic image and is suitable for execution on a standard Graphics Processor Unit. The second proposed algorithm represents an inverse mapping rendering technique which allows a simple and accurate re-sampling of synthetic pixels. Moreover, multiple techniques for properly handling occlusions are introduced. The chapter concludes with a quality evaluation of the obtained synthetic images when using the proposed rendering techniques.

 4.1 Introduction
  4.1.1 Review of image rendering techniques
  4.1.2 Acquisition and rendering using N-depth/N-texture
  4.1.3 Compression of N-depth/N-texture
 4.2 Depth Image Based Rendering
  4.2.1 Rendering using 3D image warping
  4.2.2 Rendering using triangular meshes
  4.2.3 Rendering using relief texture mapping
  4.2.4 Rendering using inverse mapping
 4.3 Occlusion-compatible scanning order
  4.3.1 Occlusion-compatible scanning for rectified images
  4.3.2 Occlusion-compatible scanning for non-rectified images
 4.4 Occlusion handling
  4.4.1 Padding using occlusion-compatible scanning order
  4.4.2 Padding using occlusion-compatible scanning order and image pre-processing
  4.4.3 Occlusion handling using multiple source images
 4.5 Experimental results on rendering quality and evaluation
 4.6 Conclusions