Chapter 14: Image-Based Rendering
Synthesizing novel views from images using view interpolation, light fields, video textures, and modern neural rendering techniques.
Chapter Overview
What if we could render new views without building explicit 3D models? Image-based rendering synthesizes novel viewpoints directly from captured images, blurring the line between photography and computer graphics.
What is this chapter about? We explore techniques for synthesizing new images from existing ones: interpolating between viewpoints, capturing complete light fields, creating seamless video loops, and using neural networks for view synthesis.
Why does this matter? Image-based rendering powers:
- Virtual tours: Walk through spaces from photographs
- 3D photos: The depth effect on social media platforms
- Visual effects: Creating impossible camera moves in film
- Telepresence: Realistic video conferencing with viewpoint changes
How the topics connect: We start with view interpolation—synthesizing views between captured images. Light fields capture all rays, enabling arbitrary viewpoint selection. Video textures create infinite loops from finite video. Finally, neural rendering (NeRF) learns to render new views from sparse inputs.
Chapter Roadmap
Click any topic to jump in
View Interpolation
Synthesizing novel views between captured images using depth-based warping — the classical approach to view synthesis.
Light Fields
Capturing all rays in a scene for post-capture refocusing and viewpoint change — complete 4D ray sampling.
Temporal and neural approaches
Video Textures
Creating infinite seamless video loops by finding natural transition points — temporal image-based rendering.
Neural Rendering
Deep learning for photorealistic view synthesis from sparse inputs — from NeRF to real-time Gaussian Splatting.
Sign up to unlock this chapter
This chapter is part of PixelBank Premium. Create a free account, then upgrade to read the full lesson — concepts, walkthroughs, and exercises.