Speed-up Feature Detector using Adaptive Accelerated Segment Test

ABSTRACT

Image matching using local features of an image patch is a primary stage for various higher level computer vision applications such as object recognition, motion tracking, simultaneous localization and mapping, stereo vision, and so on. A feature from accelerated segment test (FAST) has shown to be a superior feature detector in terms of computational speed. In our previous research, we presented a feature detector that improves the speed of FAST. In this study, we extend it to speed up it further. Experimental results on benchmark data-sets reveal that the new method speeds up FAST detector better than our previous detector with marginal improvement in repeatability score. A comprehensive review of related feature detectors is also provided to place our work in context.

KEYWORDS:

• Accelerated segment test
• Corner detection
• Feature extraction
• Image matching
• Multiscale image analysis
• Pyramid data structure
• Repeatability score

Notes

1. See http://www.robots.ox.ac.uk/∼vgg/data/data-aff.html

2. See http://www.robots.ox.ac.uk/∼vgg/research/affine

Additional information

Funding

This research was supported by the Basic Science Research Program through the National Research Foundation (NRF) of Korea funded by the Ministry of Science, ICT, and Future Planning (MSIP) [grant number NRF-2014R1A1A3051169], and also by the MSIP, under the framework of International Cooperation Program [grant number NRF-2015K2A1A2070917].

Notes on contributors

Yenewondim Biadgie

Yenewondim Biadgie received his PhD in Computer Engineering in 2012 from Ajou University, South Korea. He worked as a lecturer in Arbaminch University from 2002 to 2003 and Addis Ababa University from 2007 to 2008. He is currently a post-doctoral researcher at visual computing research laboratory of Computer Engineering Department, Ajou University. His research interest includes image and video processing, pattern recognition, computer vision, machine learning, and data mining.

E-mail: [email protected]

Kyung-Ah Sohn

Kyung-Ah Sohn received her PhD in Computer Science from Carnegie Mellon University, USA, in 2011. She was with the College of Medicine, Seoul National University, as a research professor in 2012. She is currently an assistant professor in the Department of Information and Computer Engineering, Ajou University, South Korea. Her research interest includes machine learning and data mining approaches.

E-mail: [email protected]