TQTQliu / ET-MVSNet

[ICCV 2023] When Epipolar Constraint Meets Non-local Operators in Multi-View Stereo
MIT License
62 stars 1 forks source link

Why Epipolar When we have Homography? #3

Closed AsDeadAsADodo closed 12 months ago

AsDeadAsADodo commented 12 months ago

Thanks for the work! I read several works about Epipolar Transofomer, e.g. MVSTER. Homography matrix is able to directly project point p to corresponding p'. Then why apply Epipolar Geometry to find point (p)'s corresponding epipolar line? I've read MVSTER code, they still construct a cost volume from Homography warping.

TQTQliu commented 12 months ago

Hi, constructing a cost volume through homography transformation is a part of the pipeline of cost volume-based MVS methods. Our method follows this pipeline as well, but focuses on its first phase, the feature encoding part. This paper aims to aggregate information of corresponding epipolar lines to enhance feature representation. Therefore, we extend the point-to-point homography transformation to the correspondence between the point and the epipolar line, and design the epipolar pair search algorithm. You can find these in the framework (Fig. 4) of the paper.

AsDeadAsADodo commented 12 months ago

Hi, constructing a cost volume through homography transformation is a part of the pipeline of cost volume-based MVS methods. Our method follows this pipeline as well, but focuses on its first phase, the feature encoding part. This paper aims to aggregate information of corresponding epipolar lines to enhance feature representation. Therefore, we extend the point-to-point homography transformation to the correspondence between the point and the epipolar line, and design the epipolar pair search algorithm. You can find these in the framework (Fig. 4) of the paper.

Thanks for your quick reply! So Epipolar is a way to enhance the feature maps, generate a new feature map for each one of them? After this phase, we still construct cost volume through homograhpy transformation?

TQTQliu commented 12 months ago

Yes, you're right.