This paper concerns the problem of wood cellular structure image registration. Given the large variability of wood geometry and the important changes in the cellular organization due to moisture sorption, an affine-based image registration technique is not exhaustive to describe the overall hygro-mechanical behaviour of wood at micrometre scales. Additionally, free tools currently available for non-rigid image registration are not suitable for quantifying the structural deformations of complex hierarchical materials such as wood, leading to errors due to misalignment. In this paper, we adapt an existing non-rigid registration model based on B-spline functions to our case study. The so-modified algorithm combines the concept of feature recognition within specific regions locally distributed in the material with an optimization problem. Results show that the method is able to quantify local deformations induced by moisture changes in tomographic images of wood cell wall with high accuracy. The local deformations provide new important insights in characterizing the swelling behaviour of wood at the cell wall level.

中文翻译：

---

一种非刚性配准方法，用于分析木质细胞壁中的局部变形。

本文涉及木质蜂窝结构图像配准问题。鉴于木材几何形状的大变化性以及由于水分吸收而引起的细胞结构的重要变化，基于仿射的图像配准技术并不能详尽地描述木材在微米尺度上的总体吸湿力学行为。另外，当前可用于非刚性图像配准的免费工具不适用于量化复杂的分层材料（例如木材）的结构变形，这会由于未对准而导致错误。在本文中，我们将基于B样条函数的现有非刚性注册模型改编为我们的案例研究。如此修改后的算法将局部识别在材料中局部分布的特定区域内的特征识别概念与优化问题结合在一起。结果表明，该方法能够高精度地量化木材细胞断层图像中水分变化引起的局部变形。局部变形为表征木材在细胞壁水平的溶胀行为提供了新的重要见解。