On image matching and feature tracking for embedded systems: a state-of-the-art

Pissaloux, E.E. and Maybank, Stephen J. and Velázquez, R. (2013) On image matching and feature tracking for embedded systems: a state-of-the-art. In: Chatterjee, A. and Nobahari, H. and Siarry, P. (eds.) Advances in Heuristic Signal Processing and Applications. Berlin, Germany: Springer, pp. 357-380. ISBN 9783642378799.

Abstract

This chapter presents a state-of-the-art on image and feature matching in 2D and 3D. Only methods suitable for embedded or wearable real-time system implementation are considered. The implementation may be supported by a dedicated VLSI system. Heuristic guided predictive approaches to image matching are classified as area-based or feature-based. Correlation-based matching, Fourier matching, and mutual information approaches are area-based. Graph, series, and their combinations, including pyramidal or multiresolution algorithms, are feature-based. First, relaxation, maximal clique, tree search, region growing, and dynamic programming methods are briefly described. Next, the correlation-based methods, with a fixed size or adaptive sized window, pyramidal methods, the iterative closest point (ICP) algorithm, and probability (saliency)-based approaches are sketched. Some hardware architectures which support these methods offer new computational models for image matching and image processing. Methods for feature tracking are split into two classes: correlation-based methods and Bayesian methods. Kanade–Lucas–Tomassini (KLT), three-steps/new-three-steps, four-steps, diamond efficient search, and some of their new extensions with inertial data represent the first class, while Kalman and other filters, and their recent improvements represent the second class. The importance of matching is attested by the wide number of applications which include robot navigation, navigation assistance for impaired people, navigation in virtual systems, the processing of medical, satellite and urban imagery, human computer interaction, stereo vision, 3D reconstruction, multimodal fusion. processing, remote sensing, etc.

Metadata

Statistics

DownloadsShow export options

Export as XMLJSONCSV

Activity Overview

6 month trend

0Downloads

6 month trend

243Hits

Additional statistics are available via IRStats2.

Archive Staff Only (login required)

Edit/View Item