Innovations
image_only
text_background
Lightbulb
text_background

Micro-CV Subpixel Phase-Base Image Registration

Categories for this invention

Computer Vision

Imaging

 

Intellectual Property

IMAGE REGISTRATION METHOD

PCT

WO 2017/200395

 

Image Gallery

 

CSFR1 Binding_0.png

Questions about this Technology?

Contact: Maria-Jose Alvarez

 

Summary

Efficient, accurate and robust suite of tools for the future of Computer Vision

 

Problems Addressed

Many image/signal processing tasks can benefit from very accurate registration of two or more images or signals. However, technical and practical challenges have limited the accuracy of current registration methods. Despite widespread potential applications for accurate image registration methods, they are rarely meaningfully used as current algorithms require complex and time-consuming computations, and are not sufficiently accurate. Furthermore, they often fail when no distinguishing features exist in the images.

 

Applications

  • Non-destructive tests and predictive maintenance
  • Remote monitoring
  • Security
  • Medical devices
  • Agriculture
  • Automotive Industry

 

Technology

The novel registration algorithm can provide very accurate image registration and displacement measurement to less than 0.0002 of a pixel (voxel) close to real time. This translates to a 50x more precise, and 10x faster measurement than existing state-of-the-art algorithms. Low complexity and computation demands of this algorithm allow its integration on low-end and embedded processors. Furthermore, this novel algorithm is robust, being able to measure displacements in images with low-features and poor lighting.

 

Advantages

  • Accurate and robust to find fine deformations (as small as 0.0002 pixel/voxel)
  • Efficiency allows near real-time implementation
  • Performs well in low-feature images
  • Performs well in poor lighting conditions
  • Accurate 3D reconstruction of the geometry of objects
  • Allows accurate 2D and 3D tracking of objects
  • Simultaneous calibration of a large number of cameras

PROJECT#34488

Inventors

Professor Poul Nielsen

Auckland Bioengineering Institute

Faculty of Engineering

 

Professor Martyn Nash

Auckland Bioengineering Institute

Faculty of Engineering

 

Associate Professor Andrew Taberner

Auckland Bioengineering Institute

Faculty of Engineering

 

Dr Amir Haji Rassouliha

Auckland Bioengineering Institute

Faculty of Engineering