Image Harmonization

Image harmonization aims to seamlessly integrate foreground and background elements in composite images, addressing inconsistencies in lighting, color, and style. Recent research focuses on developing unsupervised and source-free methods, employing architectures like normalizing flows, generative adversarial networks (GANs), and latent diffusion models to achieve this, often incorporating semantic segmentation for more precise control. This field is crucial for improving the quality of medical images (e.g., MRI, CT) from diverse sources, enhancing the reliability of downstream analyses, and also finds applications in image editing and artistic compositing.