Pyramidal optical flow method-based lightweight monocular 3D vascular point cloud reconstruction

Abstract

We propose a method for reconstructing a 3D point cloud of the organ model based on optical flow and take the 3D cardiovascular model reconstruction as an example. This optical-flow distribution based 3D point cloud reconstruction method is divided into four steps. Firstly, we employ the Coey filter to remove the noise points and improve the resolution of the raw images. Secondly, we implement the Shi-Tomasi method to extract the feature points from these filtered images. Thirdly, we remove the redundancy in the feature point set by the optical flow distributions. Finally, we converted the obtained feature points from 2D to 3D through the optical flow distribution and then reconstructed a 3D point cloud of the medical organ. With the help of our 3D representation, doctors and patients can view the 3D medical models on the Web. The final result on the Web shows the proposed method is feasible and superior.