Extremely Dense Point Correspondences using a Learned Feature Descriptor

TL;DR

This paper introduces a self-supervised dense descriptor learning method that improves 3D reconstruction quality from endoscopy videos by better matching points on texture-scarce surfaces, generalizing across patients and scopes.

Contribution

We propose a novel self-supervised training scheme and loss for dense descriptor learning that enhances point correspondence in endoscopic 3D reconstruction.

Findings

Our dense descriptor outperforms recent local and dense descriptors on sinus endoscopy data.

It generalizes well to unseen patients and scopes, improving SfM model density and completeness.

Effective on public dense optical flow and SfM datasets.

Abstract

High-quality 3D reconstructions from endoscopy video play an important role in many clinical applications, including surgical navigation where they enable direct video-CT registration. While many methods exist for general multi-view 3D reconstruction, these methods often fail to deliver satisfactory performance on endoscopic video. Part of the reason is that local descriptors that establish pair-wise point correspondences, and thus drive reconstruction, struggle when confronted with the texture-scarce surface of anatomy. Learning-based dense descriptors usually have larger receptive fields enabling the encoding of global information, which can be used to disambiguate matches. In this work, we present an effective self-supervised training scheme and novel loss design for dense descriptor learning. In direct comparison to recent local and dense descriptors on an in-house sinus endoscopy dataset, we demonstrate that our proposed dense descriptor can generalize to unseen patients and scopes, thereby largely improving the performance of Structure from Motion (SfM) in terms of model density and completeness. We also evaluate our method on a public dense optical flow dataset and a small-scale SfM public dataset to further demonstrate the effectiveness and generality of our method. The source code is available at https://github.com/lppllppl920/DenseDescriptorLearning-Pytorch.