We present a novel approach for registration of multimodal technical images associated with painted artworks. We developed both an automated method for area-based registration and mosaicking based on mutual information, and a graphical user interface (GUI) that allows heritage end-users to implement our method from the browser of any computer, without the need of interfacing with the code directly.
Our method was initially developed for registering element distribution maps resulting from macro X-ray fluorescence (MA-XRF) scanning of painted artworks, which take the form of a layered image stack. This stack is treated as the moving image for registration to the target fixed image - which is usually, but not limited to, the visible image of the same artwork. Our method can register multiple moving images simultaneously (each one arranged as a layered image stack, and each covering part of the fixed image), as well as it can be applied for registering various other image modalities.
Operating in consecutive stages, the method registers one moving image at a time, progressively evaluating a range of transformations that are applied to each image in the stack. Finally, a unique, optimised transformation that provides the highest average mutual information across all images in the stack is identified with consensus. This transformation is then applied to the entire moving image to obtain the best alignment between the moving and fixed images when overlapped.
Our browser-based GUI facilitates setting up the images to align, as well as storing the initial settings assigned by the user to each set of moving images. The user can then execute, automatedly, the registration of multiple moving images in the same interface - immediately after setting up initial parameters or at a later session. The results are accessible from the GUI and can be saved in the computer of the user.