Title: A similarity metric for matching incomplete edge curves

Authors: Yong Li; Robert L. Stevenson

Addresses: Department of Electrical Engineering, University of Notre Dame, IN 46556, USA ' Department of Electrical Engineering, University of Notre Dame, IN 46556, USA

Abstract: This paper presents a similarity metric for use in comparing incomplete edge curves extracted from multimodal images. A common problem with multimodal image registration based on extracted integral curves is that a curve in one image may not be exactly mapped to a corresponding curve in a different image. To address this problem, this work proposes a similarity metric based on extracted subcurves as a base primitive. Given a test and reference image, curves are extracted from both images. Junctions are detected on test curves, and used to partition the test curves into subcurves. Partitioning allows for the test subcurves being matched to a reference curve if they correspond to each other. A test subcurve is defined to be matched to a reference curve if its average distance to the reference curve is small. The number of matched curves and the number of pixels on matched curves are shown to be a useful similarity metric. Experimental results show that the presented similarity metric gives more robust and reliable results.

Keywords: multimodal images; image registration; curve matching; similarity metrics; incomplete edge curves; image processing.

DOI: 10.1504/IJMISSP.2014.065688

International Journal of Machine Intelligence and Sensory Signal Processing, 2014 Vol.1 No.2, pp.153 - 173

Available online: 03 Nov 2014 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article