Computing Reviews
Today's Issue Hot Topics Search Browse Recommended My Account Log In
Review Help
Search
Fully automatic registration of image sets on approximate geometry
Corsini M., Dellepiane M., Ganovelli F., Gherardi R., Fusiello A., Scopigno R. International Journal of Computer Vision102 (1-3):91-111,2013.Type:Article
Date Reviewed: Aug 14 2013

With the development of data storage and acquisition hardware, we will be seeing more and more applications of photo-realistic 3D models. But we are not quite there yet, as the authors point out. Existing direct acquisition techniques are limited to sophisticated situations, so the authors take a different approach, generating the 3D textured data by automatically composing a plain 3D model and a set of digital photographs.

The approach involves two steps: the so-called structure from motion (SfM) reconstruction, and 2D to 3D registration. Note that neither of these topics is new, but when they are applied to this specific area, special care must be taken. The paper focuses on this, answering tough questions while pursuing hands-on work step-by-step. If the reader is new to geometric processing or 3D reconstruction from 2D images, this paper could be a good starting point, thanks in part to the numerous references included. It is also good to see how traditional computer graphics algorithms can be concatenated to help each other, such as the combination of the variational shape approximation (VSA) algorithm and the 4-point congruent set (4PCS) algorithm. While 4PCS can be perfectly applied to 3D shape alignment, it suffers from irregular sampling of the 3D models. At this point, VSA provides a good basis for resampling the models and thus improves the confidence of 4PCS.

When considering such combinations, it is crucial to consider the tradeoff between performance and complexity. This is also true in many other cases, especially given that 3D datasets can be both huge and noisy at the same time. The authors successfully address this problem by either adding remedial algorithms or a modified version of the original algorithm. Readers who wish to try the algorithm should note that, while implementation may be straightforward, tuning the parameters can be tedious and is beyond the scope of this paper. Unless it can be verified through more experiments, readers shouldn’t expect full automation from the algorithm on any dataset. This topic will be revisited when new ideas and hardware become available.

Reviewer:  Chang Liu Review #: CR141463 (1310-0936)
Bookmark and Share
  Editor Recommended
Featured Reviewer
 
 
General (I.4.0 )
 
 
Computer Vision (I.5.4 ... )
 
 
Registration (I.4.3 ... )
 
 
Scanning (I.4.1 ... )
 
Would you recommend this review?
yes
no
Other reviews under "General": Date
Matrix structured image processing
Dougherty E., Giardina C., Prentice-Hall, Inc., Upper Saddle River, NJ, 1987. Type: Book (9789780135656235)
Jul 1 1988
Digital image processing (2nd ed.)
Gonzales R., Wintz P., Addison-Wesley Longman Publishing Co., Inc., Boston, MA, 1987. Type: Book (9789780201110265)
Jul 1 1988
Art and design: AI and its consequences
Howard G., John Wiley & Sons, Inc., New York, NY, 1986. Type: Book (9780471909309)
Dec 1 1987
more...

E-Mail This Printer-Friendly
Send Your Comments
Contact Us
Reproduction in whole or in part without permission is prohibited.   Copyright 1999-2024 ThinkLoud®
Terms of Use
| Privacy Policy