Abstract: Super-resolution reconstruction produces one or a set of high-resolution images from a set of low-resolution images. In the process of reconstruction methods, sub-pixel registration among the set of low-resolution images is a very important step. A sub-pixel registration method based on Support Vector Machine (SVM) was proposed: the related motion parameters were considered as the target set of SVM. After the mapping between the image features and the target set was built through SVM training, the motion parameters could be calculated using SVM regression. The experimental results confirm the effectiveness of the proposed method compared to other sub-pixel registration methods.

Key words: Support Vector Machine (SVM), sub-pixel registration, super-resolution, image reconstruction

摘要： 超分辨率重建是根据场景的一组低分辨率图像重建其高分辨率图像。重建算法中，低分辨图像之间的亚像素配准是很重要的一部分。提出了一种基于支持向量机的亚像素配准方法，将低分辨图像之间的相对旋转平移参数看成支持向量机的目标集，通过支持向量回归建立图像特征与目标集之间的映射关系，从而计算图像间的相对运动参数。实验表明，与现有算法相比，所提出的算法具有较高的精度。

关键词: 支持向量机, 亚像素配准, 超分辨率, 图像重建