对数极坐标特征指导的迭代就近点法

Log-polar Feature Guided Iterative Closest Point Algorithm

在多模态图像,有光照、取向及尺度或纹理变化的图像,以及宽基线图像之间通常存在大的形貌畸变。然而,当前主流推广的双自举迭代就近点法(GDB-ICP)在配准这类图对时存在困难。主要原因是,该方法利用提取的尺度不变泡点(SIFT keypoint)来引导迭代就近点法(ICP),但在大畸变图像上提取的SIFT点是不可靠的。为此,提出了一种用对数极坐标特征点(LPF)来引导迭代就近点的图像配准新方法(LPF-ICP)。实验结果表明,LPF-ICP方法成功地从Rensselaer数据组中的所有22对挑战性图对提取了可靠的LPF种子,并顺利实现了全图配准,而GDB-ICP方法则只完成了其中的19对,从而证实了LPF-ICP方法的有效性。

Images with lighting variations,rotation/optical zoom,physical changes of scene or at widely different viewpoints,can substantially change their appearance and shape when they are acquired using different modalities.Even with the state-of-the-art technology,e.g.,the generalized dual-bootstrap iterative closest point(GDB-ICP)method,it is still difficult to register those challenging images.The reason is that the GDB-ICP method uses the scale-invariant blub points(or SIFT keypoints)to drive the iterative closest point method(ICP).However,the SIFT keypoints cannot be reliably extracted from images with large appearance changes.To handle this issue,this paper proposed a novel log-polar feature guided iterative closest point(LPF-ICP)algorithm for image registration.The experimental evaluation illustrates that the LPF-ICP method can reliably extract the log-polar feature points and successfully register all the 22 image pairs contained in the Rensselaer dataset,while the GDB-ICP method only succeeds in 19 of them,thus verifying the effectiveness of the proposed method.