Auto-Calibration, Reconstruction and Assessment of Clinical Lesions from Endoscopic Image Sequence

• Main Authors: Ta-Yuan Chien, 簡大淵
• Other Authors: Yung-Nien Sun
• Format: Others
• Language: zh-TW
• Published: 2002
• Online Access: http://ndltd.ncl.edu.tw/handle/cnhr8a

碩士 === 國立成功大學 === 資訊工程學系碩博士班 === 90 === 　　In the last 30 years, the progresses in optical engineering, computer science and electronic techniques have made the endoscopy an invaluable tool in both internal clinics and surgical operations. As its applications increase exponentially, it has even become a specialized division in the clinical medicine. 　　The image analysis technique provides important aids to the processing of clinical endoscopic images. However, traditional image analyses emphasize the 2-D image distortion calibration and assessment for endoscopic images. In this thesis, we use the computer vision algorithm to reconstruct the 3-D model from the endoscopic image sequence, texture mapping with real images are then employed to enhance the visualization of the reconstructed tubular scene. 　　For obtaining a larger field of view inside a small and narrow pipeline, the endoscope is usually equipped with wide-angle lens. Therefore, the acquired images are often with certain degrees of shape distortion. Before 3-D reconstruction, our system provides a fast mechanism for correcting the wide-angle lens distortion. Using a calibration pattern, the nonlinear distortion is corrected with a simple mathematic model for the endoscopic images. Once the endoscopic lens is calibrated, the same calibration parameters can be utilized repeatedly for the calibrated instrument. 　　On the other hand, how to extract and track the correspondent features from the image sequence is one of the most important tasks in 3-D reconstruction. Our systems use the high-pass filter to extract the edge feature and the Kanade-Lucas-Tomasi (KLT) feature tracking algorithm to obtain the feature correspondences. The color information, zoom-out characteristic and distortion factor of endoscope image sequence are all taken into account for improving the feature tracking results. 　　Thereafter, the multiple frame auto-calibration is used to obtain the camera parameters. The 3-D coordinates of the detected feature points are then computed from the multiple images to reconstruct the 3D scene inside the tubular structure. At last, texture mapping with real endoscopic images is adopted to visualize the realistic 3D scene inside the reconstructed tubular structure of the observed organ.