Polyp shape estimation from endoscopy video using EKF monocular SLAM with SFS model prior

摘要

In this paper, a new approach to recover 3D shape of a polyp from the endoscopic video is proposed. For this, a point light source of illumination, perspective projection and six-degree motion of the endoscope are considered. Previous approaches attempted to recover polyp shape with the help of extended Kalman filter (EKF) and monocular simultaneous localization and mapping (SLAM) from a static scene. Our approach uses iterative data acquisition approach with the existing EKF Monocular SLAM and the intensity model prior. Our method constructs a dense 3D model of polyp by utilizing sparse data, which are collected by a Monocular EKF SLAM and a base model prior. Experimental results show the efficacy of our proposed method.