A Feature Point Extraction Method for Capsule Endoscope Localization

Wireless capsule endoscopy (WCE) has emerged as a popular non-invasive imaging tool for inspection of human Gastrointestinal (GI) tract. If a displacement technique based entirely on visual features is used for WCE positioning, a suitable visual feature extraction technique is important. In this paper, an improved ORB algorithm is proposed to extract feature points from capsule endoscopy images. Because of the complexity of the scene in the digestive tract and the insignificant image variation, the adaptive threshold is proposed to be calculated using the coefficient of variation in the feature point extraction stage, which improves the ability of the algorithm to extract feature points in homogeneous area features. Then, the feature points are further filtered using the quadtree method to eliminate over-concentration and overlapping feature points. In the feature point description phase, BEBLID is used to enhance the saliency of the feature description. Finally, the Hamming distance is used to match points and RANSAC is used to avoid mismatches. The experimental results show that the improved algorithm has better stability and adaptability to capsule endoscopic images, and effectively improves the matching accuracy on the basis of satisfying the real-time performance.