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 stitching unevenly weak texture backgrounds represented by sea-sky background images with large parallax, a multi-level composite feature extraction and optimal seam search algorithm combined with information entropy is proposed. Firstly, the images 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 neighborhood 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 algorithm 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 indicators, and visual effects of the stitching results compared to the comparison algorithm.
2024,46(19): 127-131 收稿日期：2023-12-5
DOI：10.3404/j.issn.1672-7649.2024.19.022
分类号：TP391.7
作者简介：王冠华(1995-),男,硕士研究生,研究方向为数字图像处理
参考文献：
[1] 智能船舶规范(2023) [S]. 中国船级社, 2023.
[2] 潘俊旭. 舰船监控图像拼接与识别研究[J]. 舰船科学技术, 2019, 41(24): 172-174.
PAN Junxu. Research on mosaic and recognition of ship surveillance images[J]. Ship Science and Technology, 2019, 41(24): 172-174.
[3] 李晓明, 郝沙沙, 陈双慧. 结合先验知识的海底图像配准方法[J/OL]. 计算机辅助设计与图形学学报: 1-8[2023-10-0]. http://kns.cnki.net/kcms/detail/11.2925.TP.20230815.1538.016.html.
[4] 代家印, 王育昕, 袁杰. 低空航拍全景图像拼接研究[J]. 南京大学学报(自然科学), 2023, 59(2): 239-246.
[5] LIN C C , PANKANTI S U , RAMAMURTHY K N , et al. Adaptive as-natural-as-possible image stitching[C]//Computer Vision & Pattern Recognition. IEEE, 2015.
[6] 丛一之. 基于特征增加和多项优化的图像拼接方法[D]. 长春：吉林大学, 2023.
[7] YIN H, GONG Y, QIU G. Side window guided filtering[J]. Signal Processing, 2019, 165.
[8] NIE L, LIN C, LIAO K, et al. Depth-aware multi-grid deep homography estimation with contextual correlation[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2022, 32(7): 4460–4472.
[9] 罗永涛, 张红民. 基于最佳缝合线与灰度均值差改正比的图像拼接算法[J]. 激光杂志, 2018, 39(12): 42-46.
[10] 樊逸清, 李海晟, 楚东东. 使用线约束运动最小二乘法的视差图像拼接[J]. 中国图象图形学报, 2019, 24(1): 23-30.
