A robust deformed image matching method for multi-source image matching

Abstract

Multi-source image matching is a challenging task due to the presence of image distortion, as well as significant intensity changes between image pairs in corresponding regions. In addition, the influences of variant scales and multiplicative noises will also have an adverse effect on the matching accuracy. In this paper, a combination of feature descriptor called “histogram of angle and maximal edge orientation distribution” (HAED) is proposed for multi-source image matching. First, the contour segment feature, which extracts the image information using both the angle and edge orientation distribution, presents the accurate correspondence between multi-source images. Second, the similarity calculated by using Fréchet distance metric between curves is defined as a weight parameter of each contour segment histogram to improve the matching performance. Finally, a precise bilateral matching rule is used to perform the matching between the corresponding contour segments. Infrared–visible image data sets in different environments are used for experiments. The results demonstrate that the proposed algorithm achieves a more accurate matching performance than other multi-source image matching algorithms.