Texture features for affine registration of thermal (FLIR) and visible images

The aim of our research is to analyse the importance of texture information for affine registration of gray-scale far-infrared (FLIR) images and gray-scale images taken in the visible spectrum. Texture features are extracted by Laws texture coefficients and used for computing registration criterion functions. The proposed feature based approach is compared to the commonly used approach, where a registration criterion function is computed directly from intensity features, i.e. grey values. The experiments on a small database of ten image pairs show that our texture feature based registration method works well for the given image modalities. Furthermore, we can expect more robust and more correct registration when texture based criterion function instead of intensity based one is used.