Projects::Cubic VR Movie

Cubic VR is everywhere now. One of the example the Street View mode in Google Maps. This is a computer vision project that using only one camera to create Cubic VR.

As Project::Feature Matching shows, SIFT can be used to identify the features in images. These feature points can then be used as tie points among a set of images. With tie points, it is possible to mosaic these images.

The resulting mosaics will then be treated as texture of surfaces of a box. And finally a Cubic VR is generated.

Input

source images (click here to launch), taken in Georgia Tech bookstore 2F.

Output

Quick VR (click here to download, 258KB, require Quicktime player)

Code and Deliverables

ZIP file (Code, source images, and everything else you need to run the code)

Run-Code Description (MSword file)

WriteUp I (computing Homographies)

WriteUp II (RANSAC)

Intermediate results

mosaics (click to view in full resolution)

face #1 (composed of 5 images)

face #2 (composed of 5 images)

face #3 (composed of 5 images)

face #4 (composed of 5 images)

Cropped square images (click to enlarge)

Links:

Original project page

Assignment Instruction

CS7495 2005 Class Website