【Homography Estimation】《Deep Image Homography Estimation》

arXiv-2016

---

---

1 Background and Motivation

单应性Homograph估计：从传统算法到深度学习

用卷积直接回归单应性矩阵（transformation estimation，homography estimation），8个自由度

The homography is an essential part of monocular SLAM systems in scenarios such as:

• Rotation only movements
• Planar scenes
• Scenes in which objects are very far from the viewer

2 Related Work

无

3 Advantages / Contributions

利用卷积神经网络学四个点的偏移来进行 Homography Estimation

4 Method

（1）The 4-point homography parameterization

单应性矩阵把图 $(u,v)$ 映射成了 $(u^{\prime },v^{\prime })$

H11 H12 H21 H22 与旋转有关，H13 H23 和平移有关

Balancing the rotational and translational terms as part of an optimization problem is difficult

单应性矩阵中 9个参数相互组合有实际意义，没有完全解耦干净， 9 个参数共 8 个自由度，作者直接改学图 $(u,v)$ 映射成了 ( u ′ , v ′ ) (u',v') (u