基于Gaussian-Hermite矩的图像局部特征描述与匹配研究

摘要

The description of image local features has been the most important part in image registration and image retrieval. In this paper, Gaussian-Hermite moments are studied and their invariants are derived. Due to that the scale parameter does not affect their rotation invariants, a descriptor based on multi-scale Gaussian-Hermite moments is proposed. Finally, the proposed descriptor is tested in a serial experiments with the state-of-the-art methods, and the experimental results show that the proposed descriptor performs much better than the others.%图像局部特征描述一直是图像配准与图像检索中极为重要的一部分。本文深入研究Gaussian-Hermite矩及其旋转不变矩的性质，根据尺度因子不影响其旋转不变性的特点，提出基于多尺度Gaussian-Hermite矩的图像局部特征描述方法，并通过实例与现有的特征描述方法进行详细比较。实例结果表明，所提出的方法具有更强的特征描述能力。