基于单目视觉的视网膜三维重建技术研究

摘要/Abstract

摘要： 为了在视网膜疾病诊断中获得精度高、可视化好的视网膜,提出一种基于单目视觉的视网膜三维重建算法。该算法通过对两幅预处理图像的特征点提取确定匹配点对之间的对应关系,采用RANSAC算法去除误匹配点对,准确率高,用链匹配的方法使得多幅图像间的特征点匹配从而得到整体最优化结果。根据不同图像间特征点的对应关系,使用4通光束平差法确定空间中摄像机相对位置关系,使用PMVS算法实现视网膜的三维重建。实验结果表明,该算法具有很好的鲁棒性和稳定性,能较好地实现视网膜的三维重建。

关键词: 视网膜图像, 尺度不变特征变换, 链匹配, 4通光束平差法, 三维重建

Abstract: In order to obtain high accuracy and good visualization in the diagnosis of retinal diseases, a new algorithm for 3D reconstruction was proposed based on monocular vision. Through two pieces of preprocessing image feature points extraction the algorithm was used to determine the matching relationship between pairs of points and RANSAC algorithm was adopted to remove the false matching points to obtain the feature point to improve the accuracy, the feature points matching of multiple images were obtained by using the method of chain registration. According to the corresponding relationship among the feature points in different images, the relative position of the camera in space was determined by using the method of 4 pass bundle adjustment, PMVS algorithm was used to achieve three-dimensional reconstruction of the retina. Experimental results show that the algorithm has good robustness and stability, may better realize the three-dimensional reconstruction of the retina.

Key words: retinal image, scale-invariant feature transform(SIFT), chain registration, 4 pass bundle adjustment, three-dimensional reconstruction

中图分类号:

TP391.41

王玉亮, 刘飞, 沈建新, 刘贤喜. 基于单目视觉的视网膜三维重建技术研究[J]. 中国机械工程.

Wang Yuliang, Liu Fei, Shen Jianxin, Liu Xianxi. Research on 3D Reconstruction of Retina Based on Monocular Vision[J]. China Mechanical Engineering.

参考文献

[1]邓可欣. 基于超边图匹配的视网膜眼底图像配准算法[J]. 清华大学学报(自然科学版), 2014，54(5):568-574.

Deng Kexin. Retinal Image Registration Based on Hyper-edge Graph Matching[J]. Tsinghua Univ.(Sci. & Technol.), 2014，54(5):568-574.

[2]Choe T E, Medioni G, Cohen I, et al. 2-D Registration and 3-D Shape Inference of the Retinal Fundus from Fluorescein Images[J]. Medical Image Analysis， 2008, 12(2):174-190.

[3]Lin Y, Medioni G. Retinal Image Registration from 2D to 3D[C]//IEEE Conference on Computer Vision and Pattern Recognition. Anchorage, AK, 2008:1-8.

[4]Lusk J A, Nutter B. Automated 3-D Reconstruction of Stereo Fundus Images via Camera Calibration and Image Rectification[C]//22nd IEEE International Symposium on Computer-based Medical Systems (CBMS). New York:IEEE, 2009:1-7. [5]王玉亮，沈建新，廖文和，等. 基于CUDA的眼底图像快速自动配准与拼接[J]. 中国机械工程, 2013(13):1749-1754.

Wang Yuliang, Shen Jianxin, Liao Wenhe, et al. Fast Automatic Fundus Images Registration and Mosaic Based on CUDA[J]. China Mechanical Engineering, 2013(13):1749-1754.

[6]陈萌梦，熊兴良，张琰，等. 一种视网膜眼底图像增强的新方法[J]. 重庆医科大学学报, 2014, 39(8):1087-1090.

Chen Mengmeng，Xiong Xingliang，Zhang Yan, et al. A New Method for Retinal Fundus Image Enhancement[J]. Journal of Chongqing Medical University, 2014, 39(8):1087-1090.

[7]Lowe D G. Distinctive Image Features from Scale-invariant Key-points [J]. Internatinal Journal of Computer Vision, 2004, 60(2):91-110.

[8]朱利成，姚明海. 基于SIFT 算法的目标匹配和识别[J]. 机电工程, 2009, 26(12):74-81.

Zhu Licheng, Yao Minghai. Object Matching and Recognition Based on SIFT Algorithm[J]. Mechanical & Electrical Engineering Magazine, 2009, 26(12):74-81.

[9]赵烨，蒋建国，洪日昌. 基于RANSAC的SIFT匹配优化[J]. 光电工程, 2014(8):58-65. Zhao Ye, Jiang Jianguo, Hong Richang. An Optimized SIFT Matching Based on RANSAC[J]. Opto-electronic Engineering, 2014(8):58-65.

[10]Tae E C, Medioni G. 3-D Metric Reconstruction and Registration of Images of Near-planar Surfaces[C]//11th IEEE International Conference on Computer Vision. New York: IEEE, 2007:1895-1902.

[11]杨敏，沈春林. 基于对极几何约束的景象匹配研究[J]. 南京航空航天大学学报, 2004, 36(2):235-239.

Yang Min,Shen Chunlin. Study on Scene Matching Based on Epipolar Geometric Constraint[J]. Journal of Nanjing University of Aeronautics & Astronautics, 2004, 36(2):235-239.

[12]Jin H, Soatto S, Yezzi A. Multi-view Stereo Reconstruction of Dense Shape and Complex Appearance[J]. International Journal of Computer Vision, 2005, 63(3):175-189.

[13]史利民，郭复胜，胡占义. 利用空间几何信息的改进PMVS算法[J]. 自动化学报,2011,37(5):560-568.

Shi Limin, Guo Fusheng, Hu Zhanyi. An Improved PMVS through Scene Geometric Information[J]. Acta Automatica Sinica, 2011,37(5):560-568.