Image Guidance

Image guidance leverages additional information, such as text descriptions or reference images, to improve the quality and controllability of image generation and processing tasks. Current research focuses on integrating diverse guidance modalities (text, images, frequency information) within diffusion models and transformer-based architectures to achieve precise image editing, denoising, and super-resolution, often addressing limitations in existing methods like hallucination artifacts and boundary estimation. These advancements have significant implications for various fields, including medical imaging (e.g., MRI reconstruction, surgical guidance), computer vision (e.g., scene parsing, 3D avatar generation), and creative applications (e.g., image synthesis and customization).