Project P17:
Displaying high dynamic range video

Project Goal

High dynamic range images are images with more than 8 bits per color channel, which avoid the problem of over- and underexposure of images with a lot of brightness variation. Unfortunately, common displays cannot display these high brightness changes.

There have been many papers in Siggraph and elsewhere on tone mapping, i.e. converting a high dynamic range image into one suitable for display, typically by applying compression of global intensity variations while preserving local edges in some way. In the Stanford Multi-Camera Array Project, we produce high dynamic range video. The goal of this project is to tone map such a video for display.

Project Scope

If one applies any of the published tone mappers, they will work fine on individual images, but the results will undoubtedly look inconsistent across a video sequence, because local edges come and go. The right answer is probably to extend the published tone mapping algorithms to space-time "video cubes" in some natural way - looking for gradients or features in 3D spacetime. For some algorithms, this might be very straightforward; for others, less so.

Tasks

This project is very research-oriented and can be considered quite hard. It is probably not suited for everyone.

A detailed point-by-point plan should be discussed with Marc Levoy. You can, however, start by obtaining said papers about tone-mapping algorithms from the Siggraph Proceedings.

Project Status

Andrew Adams (abadams at stanford),
Eino-ville Talvala (talvala at stanford)
2 open spots

Point of Contact

Marc Levoy and Dan Morris

Midterm Report

not yet submited

Final Report

not yet submitted