基于点线配准和超像素分割的图像拼接算法  

IMAGE STITCHING ALGORITHM BASED ON POINT-LINE REGISTRATION AND SUPERPIXEL SEGMENTATION

作　　者：蒋敏 蒋品群[1] 宋树祥[1] 夏海英[1] 李加亮 Jiang Min;Jiang Pinqun;Song Shuxiang;Xia Haiying;Li Jialiang(College of Electronic Engineering,Guangxi Normal University,Guilin 541004,Guangxi,China;College of Artificial Intelligence,Zhejiang Vocational and Technical College of Posts and Telecommunications,Shaoxing 312366,Zhejiang,China)

机构地区：[1]广西师范大学电子工程学院,广西桂林541004 [2]浙江邮电职业技术学院人工智能学院,浙江绍兴312366

出　　处：《计算机应用与软件》2023年第10期250-254,267,共6页Computer Applications and Software

基　　金：国家自然科学基金项目(61762014);广西科技重大专项项目(桂科AA18118009);桂林市科学研究与技术开发计划项目(20170113-4)。

摘　　要：针对当前一些图像拼接算法中特征点数量不足、错位和重影等问题,提出一种基于点线配准和超像素分割的图像拼接算法。该算法通过引入线特征,丰富了图像特征,增加了特征点数量,并与映射关系和几何约束相结合,提高匹配精度;采用超像素分割代替传统的网格划分,保护了图像内容,提高了单应性的准确性,改善了错位和重影。主观效果和客观指标的对比实验结果表明了该算法的有效性和优越性。Aimed at the problems of insufficient feature points,misalignment and ghosting in some current image mosaic algorithms,an image stitching algorithm based on point-line registration and superpixel segmentation is proposed.The algorithm enriched image features by introducing line features,increased the number of feature points,and combined with mapping relationships and geometric constraints to improve matching accuracy.Superpixel segmentation was used instead of traditional grid division to protect image content,improve the accuracy of homography,and improve the misalignment and ghosting.The comparative experimental results of subjective effects and objective indicators show the effectiveness and superiority of the algorithm.

关 键 词：图像拼接 点线特征点 超像素 完整性 翘曲 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]