基于超分辨率重建的亚像素图像配准

针对低分辨率图像在配准过程中精度较低的问题,提出了一种基于超分辨率重建的亚像素图像配准方法。首先,对具有1至9像素位移的图像序列进行10倍降采样,获取具有0.1至0.9亚像素位移的图像序列。然后,根据图像的获取过程建立数学模型,以Bayes理论为基础,使用最大后验概率法(MAP)对亚像素位移低分辨率图像进行超分辨率重建,获取高分辨率图像。最后,使用具有亚像素配准精度的扩展相位相关法对图像进行配准。配准实验与噪声实验表明,所提方法的最大配准误差为0.03pixel,能实现对低分辨率图像的亚像素级配准,具有配准精度高、噪声抗干扰能力强等特点,可同时满足可见光图像与红外图像的高精度配准要求。

Sub-pixel image registration based on super resolution reconstruction

Considering the low precision of low-resolution image used for the registration process,a sub-pixel image registration method based on super resolution reconstruction was put forward.Firstly,10 times down-sampling was carried out in the image sequence with 1～9 pixel displacement to obtain an image sequence with 0.1～0.9 sub-pixel displacement;then,a mathematical model was built in accordance with the image acquisition process,and Maximum A-Posteriori (MAP) method was used to carry out super resolution reconstruction in low-resolution image of the sub-pixel displacement to obtain high-resolution image based on the Bayes Theory;finally,the extended phase correlation method with sub-pixel registration accuracy was applied to conduct image registration.The registration and noise experiments inclicate that the maximum registration error is 0.03 pixel.The proposed method has high registration accuracy and strong anti-noise ability.It can realize the sub-pixel registration for low-resolution image,at the same time,it can meet the requirements for high-accuracy registration of visible image and infrared image.