In this paper, we propose a novel method for image registration and fusion of Infrared Thermography (IRT) and visible light brain images. Firstly, we introduce image registration method by employing preprocessing to bring both images to a similar level followed by similarity transformation which consists of translations, rotation, and isometric scaling. To address the curved shape of the human brain within the image, a multiresolution Demons registration is carried out. Furthermore, for the image fusion method, a Multiscale Decomposition (MSD) based on Rolling Guidance Filter (RGF) with Guided and Gaussian filters is proposed to decompose base and detail layers. To fuse base and detail layers the visual saliency weight maps for both layers are created and applied. The proposed registration and fusion method are evaluated with the clinical visible light and IRT brain cortex images. Comparisons illustrate that results are well registered and fused with clear details and fewer fusion artifacts for the human visual perception.