Three-Dimensional Modeling From Endoscopic Video Using Geometric Constraints Via Feature Positioning

摘要: The endoscope is a popular imaging modality used in many preevaluations and surgical treatments, also one of the essential tools minimally invasive surgery. However, regular endoscopes provide only 2-D images. Even though stereoendoscopy systems can display 3-D images, real anatomical structure observed lesion unavailable be judged by surgeon's imagination. In this paper, we present constraint-based factorization method for reconstructing structures registered to patient, from endoscopic proposed incorporates geometric constraints tracked instrument into traditional based on frame-to-frame feature motion endoscopically viewed scene. Experiments with synthetic data demonstrate good real-scale extraction, greater accuracy than available methods. reconstruction process accomplished few seconds, making it suitable on-line applications surgeons additional shape information, critical distance monitoring warnings.