Multi-View Stereo Using Perspective-Aware Features and Metadata to Improve Cost Volume
Feature matching is pivotal when using multi-view stereo (MVS) to reconstruct dense 3D models from calibrated images. This paper proposes PAC-MVSNet, which integrates perspective-aware convolution (PAC) and metadata-enhanced cost volumes to address the challenges in reflective and texture-less regio...
Сохранить в:
| Главные авторы: | , , , |
|---|---|
| Формат: | Artigo |
| Язык: | Inglês |
| Опубликовано: |
MDPI AG
2025-04-01
|
| Серии: | Sensors |
| Предметы: | |
| Online-ссылка: | https://www.mdpi.com/1424-8220/25/7/2233 |
| Метки: |
Нет меток, Требуется 1-ая метка записи!
|