発表論文
2010.03
内視鏡カメラ映像からの腸管の 3 次元形状の復元
概要
We present a 3-D shape reconstruction of an intestine from endoscopic images. Endoscopic images are one of the very difficult images for 3-D reconstruction. It is hard to find correspondences between such images and moreover to reconstruct the shape from the images because of poor texture in endoscopic images and the peculiar camera motions of the endoscopy. In this paper, we use the matching method using the folds of intestine and remove outliers using the knowledge of the peculiar motion. We also adopt an accurate method of 3-D reconstruction from two images without the camera motion. By using 3-D reconstructed shapes obtained from different frame pairs, we compose them into the large 3-D shape. We show the effectiveness of our method by real image examples.