3D Object Construction from 2D Images

Back to all technologies
Download as PDF
64669
Many emerging visual fields, such as medical imaging and vision-assisted robotics, use techniques to reconstruct 3D objects from 2D images. Accurate and efficient image processing is very important for producing a reliable image with image reconstruction. Many reconstruction methods require a transparent object with multiple symmetries, which limit accurate reconstruction.

Purdue University researchers have developed a computer algorithm that takes one 2D image of a 3D object and reconstructs the object. This algorithm uses a minimal amount of data (one 2D image) to produce accurate and reliable reconstructions. Unlike other reconstruction methods, this algorithm minimizes the use of depth cues like shading, texture, motion, and stereo. Instead, reconstruction is performed using symmetry, planarity, and compactness. It is also possible to use this method with either transparent or opaque objects.

Advantages:
-Uses a single 2D image
-Reproduces image constructions with minimal errors
-Tolerates noise in the original image
-Requires less input information compared to previous models

Potential Applications:
-3D Object recognition
-3D scene reconstruction
-Machine vision in manufacturing, unmanned vehicles, and robots
Oct 19, 2009
CIP-Patent
United States
8,406,567
Mar 26, 2013

Jan 9, 2008
Utility Patent
United States
8,224,065
Jul 17, 2012

Jul 6, 2012
CON-Patent
United States
(None)
(None)

May 7, 2009
Provisional-Patent
United States
(None)
(None)

Oct 17, 2008
Provisional-Patent
United States
(None)
(None)

Jan 9, 2007
Provisional-Patent
United States
(None)
(None)
Purdue Office of Technology Commercialization
1801 Newman Road
West Lafayette, IN 47906

Phone: (765) 588-3475
Fax: (765) 463-3486
Email: otcip@prf.org