摘要： 针对传统的配准只考虑几何因素的不足，提出一种基于几何和光度混合仿射模型的配准迭代算法。该算法避免keren算法因为角度的泰勒级数展开所带来的误差，大幅提高配准精度，而且使图像亮度得到保持。与传统的仿射变换相比，平均平移误差精度达到了0.165个像素。采用迭代反投影算法对配准后的图像序列进行超分辨率重建。实验结果表明，超分辨率重建结果的信噪比和结构相似度有了较大提高。

关键词: 混合仿射模型, 迭代反投影, 超分辨率重建, 配准

Abstract: According to the lack that traditional registration only considers geometric factors, in order to overcome out its disadvantage, this paper puts forward a new improvement approach based on a combined affine model of geometry and photometry. This change not only avoids the error that is brought by the keren algorithm’s tailor series expansion of angle and improves the precise of image registration greatly, but also makes the illumination maintained. The average translation error is reached to 0.165 pixels compared with the traditional affine model. Iterative Back Projection(IBP) algorithm is used for super resolution reconstruction of the image sequence. Experimental result shows that super resolution reconstruction result’s signal-to-noise ratio and the mean structural similarity index of improved algorithm have great enhancement.

Key words: combined affine model, Iterative Back Projection(IBP), super resolution reconstruction, registration

中图分类号: 

• P237