大视差下海空背景图像拼接算法

Stitching algorithm for sea and air image under large parallax

针对大视差下海空背景图像为代表的一类不均匀弱纹理背景拼接任务存在的特征点稀少及配准后畸变鬼影问题,提出一种多层次复合特征提取及结合信息熵的最佳拼接缝搜索算法。首先将待拼接图像进行侧窗滤波等预处理后构建灰度图像金字塔,提取含SIFT特征在内的复合特征点,利用RANSAC算法精选特征点计算全局单应阵进行图像配准。提出一种基于局部邻域窗的能量函数,引入动态规划思想使用信息熵最大点作为拼接缝搜索起点完成拼接。试验结果表明,在自制及公开数据集UDIS-D上本文特征提取算法较常见SIFT、SURF、ORB算法平均提高122.6%、61.8%和3.75%,在拼接结果的PSNR、SSIM指标及视觉效果上本文算法较对比算法均有提升。

In view of the problems of sparse feature points and distorted ghosting after registration in the task of stitch-ing unevenly weak texture backgrounds represented by sea-sky background images with large parallax,a multi-level com-posite feature extraction and optimal seam search algorithm combined with information entropy is proposed.Firstly,the im-ages to be stitched are preprocessed by side window filtering to construct a grayscale image pyramid,and composite feature points including SIFT features are extracted.The RANSAC(random sample consensus)algorithm is used to select feature points to calculate the global homography matrix for image registration,Then,an energy function based on local neighbor-hood windows is proposed,and the dynamic programming idea is introduced to use the information entropy maximum point as the starting point for seam search to complete the stitching.The experimental results show that the feature extraction al-gorithm in this paper improves by an average of 122.6%,61.8%,and 3.75%compared to the common SIFT,SURF,and ORB algorithms on the self-made and public dataset UDIS-D.The algorithm in this paper has improved the PSNR,SSIM in-dicators,and visual effects of the stitching results compared to the comparison algorithm.