3D reconstruction of the WCE images by affine SIFT method

Abstract

Wireless capsule endoscopy (WCE) has been gradually applied for examining the gastrointestinal (GI) tract, particular the small intestine. One WCE video comprised over 50,000 images requires more than 1.5 hours in an average for the physician to give a full assessment of the patient's GI tract. However, for the power consumption reason and the miniature-design, the captured images are suffering from the resolution and illumination invariance. Moreover, the motion of the WCE camera is uncontrollable increases the complexity of the WCE images inspiration and the time needed for analyzing the image sequences even for experienced endoscopiest. Due to these limitations, a tool is required to help physicians to make fast and reliable diagnosis. In this paper, we propose a method to reconstruct the three dimension surface of the intestinal wall by the affine scale invariant feature transform (SIFT). This algorithm can help to reconstruct the GI tract's tube-like surface within selected consecutive frames. It estimates the relative movement of the WCE by applying the affine SIFT feature detector and descriptor to a sequence of WCE images. Epipolar geometry is employed to further constrain the matching feature points in order to obtain accurate 3D view. The experiments on real WCE images are presented to show the performance of our proposed method. As a conclusion, our system can provide a more informative and friendly virtualization of the WCE images to reduce the total time spent for abnormalities detections for physicians.