Non-rigid image registration for visible color and thermal IR face

Multi-sensor data fusion between visible sensors and infrared (IR) sensors plays an important role in computer vision and pattern recognition. However, the current methods using single feature based correspondence estimators with a rigid or affine transformation are unable to build accurate alignments when scenes are not planar. In this work, we present a non-rigid image registration method for visible color and thermal IR face. We evaluated the performance of the proposed method on a public database, and compared with state-of-the-art methods where our method shows the best alignments in most scenarios.