Optics and Precision Engineering, Volume. 28, Issue 9, 2085(2020)
Three-dimensional porous structure reconstruction for low-resolution monocular endoscopic images
A three-dimensional porous structure is a common application of the endoscope in the field of medical surgery and industrial inspection. However, the limitations of low endoscopic imaging resolution, weak texture features, and high noise make it difficult to utilize a traditional structure from motion (SFM) algorithm for reconstructing and locating the three-dimensional porous structure. In this study, a three-dimensional porous structure reconstruction method based on low-resolution monocular endoscopic images was proposed. The image enhancement algorithm based on weighted guided filtering is first applied to improve the endoscopic image details. The fast marching method is then used to remove floating impurities, and an outlier detection method is proposed to optimize the results of feature matching. Additionally, the sparse point cloud is obtained by an SFM algorithm. Finally, a method based on template matching and least square fitting is presented to extract the three-dimensional structures of holes. Experimental results indicated that the relative errors in the spacing between holes and their diameters of reconstruction on a porous hemisphere were <10%, and the reconstruction results on the real renal surgery video were consistent with reality. Therefore, the validity, accuracy, and feasibility of the whole reconstruction algorithm flow were verified.
Get Citation
Copy Citation Text
ZHANG Jian-xun, HAN Ming-hui, DAI Yu. Three-dimensional porous structure reconstruction for low-resolution monocular endoscopic images[J]. Optics and Precision Engineering, 2020, 28(9): 2085
Category:
Received: Apr. 7, 2020
Accepted: --
Published Online: Dec. 28, 2020
The Author Email: ZHANG Jian-xun (zhangjx@nankai.edu.cn)