A feature point extraction and matching algorithm for Wireless Capsule Endoscope image

For the Wireless Capsule Endoscopy (WCE), one of the most important medical actions is to find the size of the target disease tissue, on which the proper diagnosis can be made to determine the further medical operation. Since 2D estimation results in large error for the tissue size, 3D calculation is preferred. To obtain 3D information for the objective tissue, we can apply 3D reconstruction for the GI track and the target tissue. It can be realized by the localization and orientation of the endoscope by the calculation on the corresponding points on the successive images; and to do so one important step is to select the feature points and determine their corresponding positions. In this paper, we present an algorithm for image feature extraction and corresponding matching on the two successive endoscopic images. It is evaluated on the different area and different noises. The experimental results show that the algorithm has high feasibility and good accuracy.