Image alignment is a crucial step to generate a high quality panorama. The state-of-the-art approaches use local-adaptive transformations to deal with multi-view parallax, but still suffer from unreliable feature correspondences and high computational cost. In this paper, we propose a local-adaptive and outlier-tolerant image alignment method using RBF (radial basis function) approximation. To eliminate the visible artifacts, the input images are warped according to a constructed projection error function, whose parameters are estimated by solving a linear system. The outliers are efficiently removed by screening out the abnormal weights of RBFs, such that better alignment quality can be achieved compared to the existing approaches. Moreover, a weight assignment strategy is introduced to further address the overfitting issues caused by extrapolation, and hence the global projectivity can be well preserved. The proposed method is computationally efficient, whose performance is verified by comparative experiments on several challenging cases.

图像对齐是生成高质量全景图的关键步骤。最先进的方法使用局部自适应变换来处理多视图视差，但仍遭受不可靠的特征对应和高计算成本的困扰。在本文中，我们提出了一种使用RBF（径向基函数）逼近的局部自适应且离群的图像对齐方法。为了消除可见的伪像，根据构造的投影误差函数对输入图像进行变形，该函数的参数通过求解线性系统进行估算。通过筛选出RBF的异常权重，可以有效地去除异常值，因此与现有方法相比，可以获得更好的对齐质量。此外，引入权重分配策略以进一步解决由外推法引起的过拟合问题，因此可以很好地保留全局投影性。所提出的方法在计算上是有效的，其性能已通过在一些具有挑战性的情况下的对比实验得到了验证。