基于SIFT匹配和RANSAC算法的超分辨率重建

Super-resolution reconstruction based on SIFT matching and RANSAC algorithm

图像超分辨率重建技术是根据序列图像间信息互补重建高分辨率图像的技术,其主要步骤在于精确运动估计算法和有效超分辨率重建算法。针对存在旋转、缩放变换的序列图像,本文提出一种基于SIFT匹配和随机采样一致性算法(RANSAC)的运动估计算法,该方法首先使用SIFT算法对图像序列的特征点进行提取并匹配,然后使用RANSAC算法消除误匹配点并获取投影变换矩阵,从而获得图像序列间的亚像素级的运动信息;采用一组低分辨率序列图像进行试验,基于上述运动估计算法,采用迭代反投影进行超分辨率重建。试验结果表明,运动估计精度较高,重建影像具有较好的视觉效果,尤其适用于影像序列间存在旋转缩放运动的图像序列的超分辨率重建。

Image super-resolution reconstruction technology is a technique for reconstructing high-resolution images based on information complementation between sequence images.The main steps are accurate motion estimation algorithm and effective super-resolution reconstruction algorithm.For the sequence images with rotation and scaling transformation,this paper proposes a motion estimation algorithm based on SIFT matching and RANSAC algorithm.Firstly,the SIFT algorithm is used to extract and match the feature points of the image sequence.Finally,the random matching consistency algorithm(RANSAC)is used to eliminate the mismatched points and obtain the projection transformation matrix,so as to obtain the sub-pixel motion information between the image sequences.In this paper,a set of low-resolution sequence images are used for experiments.Based on the above motion estimation algorithm,iterative back projection is used for super-resolution reconstruction.The experimental results show that the motion estimation accuracy is high,and the reconstructed image has better visual effects,especially for the super-resolution reconstruction of the image sequence with rotational scaling motion between image sequences.