Intellectual Property Office

Non-Confidential Disclosures

“Rotation Symmetry Group Detection via Frequency Analysis of Frieze-Expansions”

PSU Inv. Disc. No 2007-3368

Field of the Invention/Keywords:

Symmetry detection, Rotation Symmetry, Frieze-Expansions, computer/machine vision

Inventors:

Y. Liu, S. Lee, R. Collins

Background:

Symmetry is ubiquitous in natural and man-made images, and symmetric figures attract human attention easily out of random natural scenes. Symmetry detection on 2D or 3D images has been an active research area for over four decades. Symmetry is a type of shape regularity. A symmetric object can be characterized efficiently by their symmetry groups, yielding a low dimensional but powerful and efficient set of features for object representation, matching, segmentation and tracking. Determining symmetry in imagery is an important application in computer vision, mathematics, architecture, art and other image processing fields. This is the field addressed by the present technology.

Invention description:

The present research shows that it is possible to detect rotational symmetry in real-world images using a novel and effective algorithm. This algorithm can detect and discriminate all three rotation symmetry groups (continuous, n-fold cyclic, and n-fold dihedral). The algorithm breaks down a 2D pattern rotation symmetry problem to a simple 1D pattern translation symmetry detection problem. A frequency analysis method is then performed on the 1D pattern, and finally a rotational symmetry strength function is used to detect all potential centers of rotation symmetry and generate a dense map of the given image. Figure 1 shows a sample image the output of the symmetry detection program.

Advantages:

• Discriminates from all 3 types of rotation symmetry groups, discrete and continuous
  
• Automatically identifies very high order of rotation symmetry (up to 90 folds or more)
  
• Computation does not depend on the order and the complexity of the symmetry detected thus efficient for complex images
  
• Uses a continuous rotation symmetry function to detect all potential centers of rotation symmetry
  
• Generates and distinguishes multiple layers of rotation symmetry with different folds
• Can detect exact symmetry region
• Fast execution
  

Contact:

Mr. Bradley A Swope
Sr. Technology Licensing Officer
Intellectual Property Office
113 Technology Center
The Pennsylvania State Univ.
University Park, PA 16802-7000
Phone: (814) 863-5987
Fax: (814) 865-3591
E-mail:bradswope@psu.edu