The Graph Camera is a tool for generating multi-perspective views of a complex 3-D environment. The Graph Camera begins as a standard planar pinhole camera frustum. That frustum then undergoes a series of bending, splitting, and merging operations. The resulting images are mostly continuous and allow for comprehensive views of 3-D space.
Bend Operation
Split Operation
Merge Operation
Graph Camera Frustum
Graph of Frusta
Navigation
One application of the Graph Camera is enhanced navigation allowing for more comprehensive views as the camera moves through the environment. Thus far, we have developed 3 versions of Graph Camera enhanced navigation.
Portal-based graph camera image (top left and fragment right) and PPC image for comparison (bottom left)
Occluder-based graph camera image (top left), PPC image for comparison (bottom left), and ray visualizations (right)
Enhanced street-level navigation
3-D Scene Summarization
Another application of the Graph Camera is 3-D Summarization. The goal of this application is to summarize a 3-D environment in a single image. Using a interactive tool, the user is able to bend, split, and merge the Graph Camera Frustum until all building in the scene are visible in the summarized image.
Graph camera image that summarizes a cartoon town scene (left) and conventional image for comparison (right)
Survellance
A final application for the Graph Camera is in survellance. The frustum of the Graph Camera is a collection of planar pinhole camera frusta, allowing the virtual frusta to be replaced with physical cameras. This allows us to create a comprehensive view for a real-world space.
Single-image comprehensive visualization of real-world scenes. The graph camera image (left) seamlessly integrates 3 video feeds (right) and shows all 3 branches of the T corridor intersection