SBP-Net: Learning Thin Structure Reconstruction with Sliding-Box Projections
2026-06-02 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors address the difficulty of reconstructing very thin 3D structures, like blood vessels or pipes, which are hard to capture because they're sparse and vary in size. They create 2D depth images from small 3D sections and use a neural network to fill in missing thin parts in these images. Then, they combine these 2D fixes back into the full 3D shape to get a detailed result. Their method works better at keeping fine details compared to previous approaches, as shown in tests on lung artery and pipeline scans.
3D reconstructionthin structureslocal depth projectionneural networkorthographic projectionpulmonary arteryCT imagingindustrial pipelineshape fusionsparse geometry
Authors
Ofir Gilad, Andrei Sharf
Abstract
Reconstructing thin 3D structures is challenging due to their sparsity, scale variation, and complex geometry. Such structures arise in a wide range of domains, including medical imaging of vascular systems and industrial pipe systems. While recent neural methods perform well on dense surfaces, they often fail to recover fine thin geometries. We propose a reconstruction approach based on local depth projections, which provide an efficient and informative 2D representation of thin structures. Specifically, we traverse the 3D model with a sliding box to generate local orthographic depth projections, which are processed by a neural network to reconstruct missing thin structures in 2D. The local reconstructions are subsequently fused back into the 3D model to produce a coherent and detailed shape. Experiments on pulmonary artery reconstruction from CT volumes and industrial pipeline recovery from synthetic and real scans demonstrate improved preservation of fine structural details over existing methods.