Chapter 10: Computational Photography
Advanced image processing techniques including HDR imaging, tone mapping, image matting, and neural style transfer that extend the capabilities of traditional photography.
Chapter Overview
Computational photography goes beyond what traditional cameras can capture. By combining multiple images or applying sophisticated algorithms, we can overcome physical limitations of camera sensors.
What is this chapter about? We explore techniques that enhance and extend traditional photography: capturing the full dynamic range of scenes, separating foreground from background, and even transferring artistic styles between images.
Why does this matter? These techniques power smartphone photography:
- HDR mode: Combining exposures for better highlights and shadows
- Portrait mode: Background blur through computational depth estimation
- Night mode: Long exposure simulation through frame stacking
- Filters: Real-time style transfer and enhancement
How the topics connect: We start with HDR imaging—capturing scenes brighter than sensors can handle. Tone mapping compresses this range for display. Image matting separates foreground from background. Finally, neural style transfer shows the creative potential of deep learning.
Chapter Roadmap
Click any topic to jump in
HDR Imaging
Capturing full dynamic range — multi-exposure fusion, camera response recovery, and weighting functions.
Tone Mapping
Compressing HDR to displayable range — Reinhard operators, white point control, and log-average luminance.
HDR pipeline: acquisition to display
Image Matting
Separating foreground from background — compositing equation, color line model, and Laplacian matting.
Neural Style Transfer
Artistic image transformation — content loss, Gram matrix style loss, and optimization-based synthesis.
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.