摘要
血细胞显微图像拼接时存在特征点冗余导致匹配效率低和过拟合的问题。基于自主搭建的血涂片自动扫描显微视觉平台,系统地研究高效、精确拼接技术。研究Forstner图像特征点检测的性能指标,发现Forstner算子具有较高准确度;针对Forstner算子检测特征点冗余的问题,提出区域匹配和图像定向滤波的复合预处理算法,减小无效搜索区域和图像噪声信息。实验结果表明:采用Forstner算子+复合预处理算法,图像拼接效果良好。对比传统全区域搜寻、匹配特征点的算法,匹配率提升58.69百分点,拼接耗时缩短了90.13%。
There are problems such as redundant feature points due to matching inefficiency and over fitting during stitching when blood cell micrographs are stitched.Based on the self-built automatic scanning micro-vision platform for blood smears,efficient and accurate stitching technology was studied systematically.The Forstner operator performance was analyzed,and it was found that Forstner operator has higher accuracy.Aimed at the redundant feature points of Forstner operator,a composite preprocessing algorithm of region matching and image orientation filtering was proposed to reduce the invalid search area and image noise.The experimental results show that the image stitching effect is good while using the Forstner operator and the composite preprocessing algorithm.Compared with the traditional full area search and matching algorithm for feature points,the matching rate is increased by 58.69%,and the stitching time is reduced by 90.13%.
作者
张从鹏
谢佳成
熊国顺
Zhang Congpeng;Xie Jiacheng;Xiong Guoshun(School of Mechanical and Materials Engineering,North University of Technology,Beijing 100144,China)
出处
《计算机应用与软件》
北大核心
2022年第9期108-112,127,共6页
Computer Applications and Software
基金
北京市教委基本科研项目(110052371803)。
关键词
医疗图像处理
计算机辅助检测
特征点检测
图像拼接
Medical image processing
Computer-aided detection
Feature point detection
Image stitching
