A Flexible Rendering Architecture for General Linear Cameras and other Multi-perspective Images

Augusto Román

CS448 - Final Project

Spring 2004

Objective

The goal of this project is to implement a flexible rendering architecture that allows rendering any of the General Linear Cameras described in Yu & McMillan’s ECCV 2004 [1] paper with the same title. The implementation should not be dependent upon specific input or output formats, favoring flexibility and extensibility over speed and efficiency.

Report

View the report as a pdf file (438 KB)

View the students.avi animation (31.2 MB)

View the students_sap.avi animation (10.2 MB)

These videos are encoded using the X-Vid codec. Here is a windows binary. (600 KB)
(it's an excellent codec!)

References

General Linear Cameras, Yu & McMillan, ECCV 04 - pdf (408 KB)

Images

This first set shows the images from the first video (students.avi)

Description	Real	Synthetic
Perspective
Orthographic
Oblique (Orthographic from the left)
Oblique (Orthographic from the bottom-left)
Twisted Orthographic
Twisted Orthographic from the bottom
Twisted Orthographic from the top
Pushbroom
Cross-slits
Cross-slits from the left
Cross-slits from the right
Pencil
Bilinear

And some frames from the second video (students_sap.avi):

Description	Image
No interpolation, no synthetic aperture, image plane facing virtual camera
Titled image plane
Partial synthetic aperture effect
Full synthetic aperture effect
Changing the image plane location