A Feature Matching Algorithm Based on an Illumination and Affine Invariant for Aerial Image Registration

This paper presents a new feature matching algorithm for registering aerial images with large illumination distortion, similar patterns, affine transformation and low overlapping areas. In this algorithm, affine invariant regions are obtained by a K-NN graph and its corresponding adjacent graph. Then an illumination and affine invariant called Illumination Invariant MultiScale Autoconvolution (IIMSA), which is the combination of MSA and MultiScale Retinex (MSR), is defined to describe the triangle regions. Based on the IIMSA, outliers are removed by a self-adaptive feature matching strategy. The performance of the proposed algorithm is evaluated by registering both visible and thermal infrared (thermal-IR) images. Compared with MSA, the accuracy of the proposed algorithm is highly improved. Copyright © 2014 IFSA Publishing, S. L.