基于人工交互的多模态图像亚像素配准

Sub-Pixel Multimodal Image Registration by Human Interaction

基于特征点自动匹配的图像配准技术通常无法实现亚像素精度的配准,在多模态图像集上甚至无法完成整像素配准.为了提高多模态图像配准精度,对亚像素图像配准技术进行研究,提出了一种基于人工交互的适用于多模态图像的亚像素配准算法.对待配准图像和参考图像输入控制点,利用投影变换和最小线性平方差算法进行粗配准,根据双边平均配准误差对控制点进行亚像素调整,从而达到精确配准.定性与定量实验结果表明,相比基于尺度不变特征和局部强度不变的特征描述符配准算法,该算法具备更高的配准精度,可显著提高多模态图像配准性能.

Image registration based on key-point mappings usually provides alignment of integer-pixel precision. Sub-pixel registration is of great challenge to the technique exploiting key-point mappings. The authors proposed an interactive algorithm to address the sub-pixel registration problem. The proposed algorithm comprises two steps,the first step is to input control points and is getting a rough registration by using projection transform and linear least square algorithm,the second step is to adjust the control points with sub-pixel step. The average distance of control points was applied to quantitatively measure registration quality. The evaluation method combined with subjective and objective judgment was used. Experiment shows that the proposed algorithm can achieve sub-pixel registration result. The performance will be more reliable than other registration technique using scale invariant feature transform and partial intensity invariant feature descriptor,and also the performance of multimodal image registration gets significantly improved.