Due to the difference of imaging mechanism, the modality difference and noise interference bring great challenges to the registration of multimodal remote sensing images. A novel multimodal remote sensing image registration method was proposed, which integrated relative total variation and phase consistency. By introducing relative total variation to eliminate ineffective textures and noise, the structural features of multimodal remote sensing images was fully extracted, and further the scale space was constructed to extract reliable feature points. Based on relative total variational scale space, the phase consistency, which is insensitive to intensity changes, was utilized to replace image gradient and construct salient feature descriptors. Experiments were conducted on three different types of datasets of multimodal remote sensing image. Compared with other state of the art registration methods, the proposed method can effectively process multimodal remote sensing images with modality differences and noise, and obtain better registration accuracy.
传统多模态遥感图像配准方法受图像噪声和非线性灰度差异的影响,提取的特征点重复率低,同时构建的描述符显著性差,大大影响了方法的配准精度。研究者发现多模态遥感图像之间存在相似的结构信息,提取图像之间稳定的结构特征可以有效降低配准任务中的模态差异和噪声干扰。传统基于高斯滤波的方法在一定程度上能够有效抑制噪声,但同时不可避免地会带来相同程度的图像模糊,从而损失大量图像边缘细节。FAN等[7]采用非线性扩散滤波构造图像尺度空间,旨在抑制噪声并保留边缘细节,深度挖掘图像内的结构信息。同样,YAO等[8]提出了一种新颖的共现滤波空间匹配(Co⁃occurrence Filter Space Matching,CoFSM)方法,并利用Sobel算子对梯度的计算进行改进,用于后续特征点提取和描述符生成,在多种遥感图像的配准任务中展现了良好的效果。然而,梯度信息对多模态遥感图像中的非线性灰度差异和对比度变化十分敏感,难以有效地描述图像中的稳定特征。ZHANG等[9]利用对灰度变化和对比度变化不敏感的相位一致性信息,建立了一种新颖的加权相位方向直方图,用于生成鲁棒的特征描述向量。FAN等[10]提出了一种自适应描述符结构,将不同尺度下的相位一致性信息进行编码,从而提高对几何形变和非线性灰度差异的鲁棒性。这些方法为多模态遥感图像配准提供了新的解决方案,但在抑制图像噪声和保持边缘细节方面仍存在局限性。
本文方法的具体流程如图2所示。首先,采用相对全变分模型为输入图像提取结构图,构造尺度空间。其次,计算相位一致性最大矩图,在最大矩图上执行FAST(Features from Accelerated Segment Test)检测算子提取数量充足的特征点。然后,根据图像的相位一致性信息为每个特征点分配主方向,并构建相位一致性方向直方图形成特征描述符。最后,采用最近邻距离比和快速样本一致性,完成特征匹配和误匹配点滤除。
采用结构稳定的类GLOH(Gradient Location and Orientation Histogram)[13]结构描述特征点,将特征区域划分为3个不同半径的同心圆,按照角度将外围2个环形区域均匀划分为8个方向区间,共获得17个图像子区域。与GLOH描述符不同的是,本文基于相对全变分尺度空间统计每个子区域的相位一致性方向直方图。综合考虑各方面的性能,为每个子区域计算一个8维相位一致性方向直方图,最终将其合并为一个136维的特征向量,形成描述符。
为验证所提方法的配准性能,在3种不同类型的多模态遥感图像数据集上进行实验,并与具有旋转不变性的RIFT[3]和CoFSM[7]两个先进方法进行比较。为定量评价方法的性能,采用图像配准方法中常用的两个评价指标:均方根误差(Root Mean Square Error, RMSE)和正确匹配点数(Number of Correct Matches, NCM)。
FANJianwei, WUYan, LIMing, et al. SAR and optical image registration using nonlinear diffusion and phase congruency structural descriptor[J]. IEEE Transactions on Geoscience and Remote Sensing, 2018, 56(9): 5368⁃5379.
LIDongchen, XIANGWenhao, DANGQiannan, et al. SAR and optical images registration using uniform distribution and structure description⁃based ASIFT[J]. Acta Geodaetica et Cartographica Sinica, 2020, 49(12): 1583⁃1590.
[4]
LIJiayuan, HUQingwu, AIMingyao. RIFT: Multi⁃modal image matching based on radiation⁃variation insensitive feature transform[J]. IEEE Transactions on Image Processing, 2020, 29: 3296⁃3310.
[5]
FANJianwei, XIONGQing, YEYuanxin, et al. Combining phase congruency and self‑similarity features for multimodal remote sensing image matching[J]. IEEE Geoscience and Remote Sensing Letters, 2023, 20: 1⁃5.
[6]
LOWED G. Distinctive image features from scale⁃invariant keypoints[J]. International Journal of Computer Vision, 2004, 60(2): 91⁃110.
[7]
XIANGYuming, WANGFeng, YOUHongjian. OS⁃SIFT: A robust SIFT⁃like algorithm for high⁃resolution optical⁃to⁃SAR image registration in suburban areas[J]. IEEE Transactions on Geoscience and Remote Sensing, 2018, 56(6): 3078⁃3090.
[8]
FANJianwei, YEYuanxin, LIUGuichi, et al. Phase congruency order⁃based local structural feature for SAR and optical image matching[J]. IEEE Geoscience and Remote Sensing Letters, 2022, 19: 1⁃5.
ZHANGYongjun, YAOYongxiang, WANYi, et al. Histogram of the orientation of the weighted phase descriptor for multi⁃modal remote sensing image matching[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2023, 196: 1⁃15.
[11]
FANJianwei, YEYuanxin, LIJian, et al. A novel multiscale adaptive binning phase congruency feature for SAR and optical image registration[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022, 60: 1⁃16.
[12]
XULi, YANQiong, XIAYang, et al. Structure extraction from texture via relative total variation[J]. ACM Transactions on Graphics, 2012, 31(6): 139.
[13]
KOVESIP. Phase congruency detects corners and edges[C]//Digital Image Computing: Techniques and Applications 2003, Sydney, 2003: 309⁃318.
[14]
GAOChenzhong, LIWei, TAORan, et al. MS⁃HLMO: Multiscale histogram of local main orientation for remote sensing image registration[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022, 60: 1⁃14.