Image-based rendering and modeling in video-endoscopy

We present algorithms for image-based rendering and modeling in video-endoscopy. Our algorithms are aimed at alleviating two common viewing problems in video-endoscopy: that the scope view is often distorted and oriented wrongly (a dis-orientation problem), and that the scope view is acquired from a viewpoint that deviates from the surgeon's (a dis-association problem). Our solutions strive to alleviate these problems to arrive at a more "open-surgery-like" and "surgeon-centered" display. We present experimental results based on real endoscopic video to illustrate our image enhancement procedures.