Non Rigid Registration

Non-rigid registration aims to align two shapes or images that have undergone non-rigid deformations, a crucial task in various fields like medical imaging and computer vision. Current research emphasizes robust and efficient algorithms, often employing deep learning architectures such as graph convolutional networks and transformers, or leveraging techniques from optimal transport theory to handle noisy, incomplete, or partially overlapping data. These advancements improve accuracy and speed in applications ranging from medical image analysis (e.g., liver surgery, radiotherapy planning) to 3D face modeling and autonomous robotic surgery, enabling more precise and reliable analyses.