Communications of the ACM
New Method Helps Computer Vision Systems Decipher Outdoor Scenes
Credit: Yelp.com
Carnegie Mellon University (CMU) researchers have developed a system that enables computers to extract more information from an image by analyzing the physical constraints of the scene. The CMU computer system analyzes an outdoor scene by using virtual blocks to build a three-dimensional model of the image that makes sense based on volume and mass.
CMU professor Alexei A. Efros says the approach could be used to understand the objects in a scene, in addition to the spaces in between those objects and what might lie behind areas blocked by objects in the foreground. The method involves separating the image into various segments that correspond to objects in the image. Once the ground and sky are identified, other segments are assigned potential geometric shapes.
The method is more than 70 percent accurate in estimating the layout of surfaces, and its performance is almost as good when comparing its segmentation to ground truth, says CMU's Abhinav Gupta.
From Carnegie Mellon University
View Full Article
Abstracts Copyright © 2010 Information Inc., Bethesda, Maryland, USA 
No entries found