DINE: Distance Is Not Enough -- Learning Global Deformation Priors for Robust Soft-Tissue Point Cloud Registration
Computer Vision and Pattern Recognition
Summary
The authors address a problem in matching soft-tissue shapes that can change a lot, which is hard because of noise and errors. They introduce DINE, a method that improves matching by adding a learned statistical understanding of how the tissue can realistically deform, not just measuring point distances. They test DINE on two existing systems and show it reduces errors and handles noise and outliers better, especially when tissue deformation is large. Their results suggest that considering the overall realistic shape changes helps improve soft-tissue point matching.
Authors
Sara Monji-Azad, Rohit Beer, Marvin Kinz, Claudia Scherl, Jürgen Hesser
Abstract
Non-rigid point cloud registration is central to soft-tissue shape analysis, but large deformations, noise, and outliers make correspondence estimation challenging. Most learning-based methods rely on local objectives such as Chamfer distance, which encourage point-wise proximity but do not constrain the global plausibility of the predicted deformation field. We address this limitation with DINE, a maximum a posteriori framework that augments distance-based registration with a learned statistical prior over displacement vector fields. DINE is applied to two registration backbones, Robust-DefReg and DefTransNet, using a two-stage strategy: a first-stage model is trained with Chamfer distance, its predicted deformation fields are used to estimate a prior, and the model is then refined with a combined distance and negative log-prior objective. We compare a full-field PCA Gaussian prior with a per-vector normalizing-flow prior. Experiments on DeformedTissue and SynBench show lower mean Chamfer distance under deformation and corruption. On DeformedTissue, DINE-PCA reduces Chamfer distance by approximately 27--69\% relative to the corresponding Stage-1 backbone across deformation levels, and improves robustness by up to 66\% for outliers and 83\% for Gaussian noise. On SynBench, improvements are modest at the smallest deformation levels and reach approximately 59--79\% from moderate to severe deformation. These results suggest that global deformation plausibility is an important constraint for reliable soft-tissue point cloud registration. (The code will be published soon.)