An Occlusion Removal Approach for Surgical Instruments Based on the Optical Flow-Guided Models

Surgical instrument removal has become increasingly necessary to provide precise information during intraoperative execution. Occlusions in the endoscopic view caused by interactions between surgical instruments and organs can impede a surgeon's ability to accurately assess internal tissue, as well as certain computer-assisted surgical algorithms that rely on endoscopic images. This work demonstrates a novel application of removing surgical instruments from endoscopic video utilizing flow-guided image inpainting methods. A comparative experiment was conducted to validate the advantages of the improved method. Additionally, two task-based experiments including target tracking and image fusion experiments were conducted to further demonstrate its potential as a preprocessing step for other visual algorithms.