用于面诊的人脸全景图像拼接算法  

Face panoramic image mosaic algorithm for face diagnosis

作　　者：宁雪丽 陈兆学[1] NING Xueli;CHEN Zhaoxue(School of Medical Instrument and Food Engineering,University of Shanghai for Science and Technology,Shanghai 200093,China)

机构地区：[1]上海理工大学医疗器械与食品学院,上海200093

出　　处：《中国医学物理学杂志》2020年第4期456-462,共7页Chinese Journal of Medical Physics

基　　金：上海市教委科研创新项目(13YZ069)。

摘　　要：采用一种基于人脸特征的柱面投影方法,将人的头部近似看做一个圆柱体,有效地解决了在采集过程中因面部角度所引起的视觉不一致性;利用SIFT特征匹配算法提取两幅图像的特征向量,并通过RANSAC匹配优化算法消除错误的匹配,实现图像的配准;接着采用渐入渐出的融合算法,使图像间实现平滑的过渡,消除拼接缝隙。实验结果表明,本研究使用的算法能够快速、有效地生成人脸全景图像,为后续中医面诊奠定了基础。A cylindrical projection method based on facial features is used to approximate the human head as a cylinder, which effectively solves the problem of visual inconsistency caused by different facial angles during image acquisition. SIFT feature matching algorithm is used to extract the feature vectors of two images, and the optimized RANSAC matching algorithm is used to eliminate false matches, thereby realizing image registration. Then fade-in and fade-out fusion algorithm is applied to smooth image transitions and eliminate stitching gaps. The experimental results show that the proposed algorithm can generate the panoramic image of face quickly and effectively, which lays a foundation for the follow-up face diagnosis in traditional Chinese medicine.

关 键 词：中医面诊 人脸全景图像 柱面投影 图像配准 

分 类 号：R318[医药卫生—生物医学工程] TP391[医药卫生—基础医学]