A capsule endoscope image classification method based on multi-feature fusion

Abstract

A capsule endoscope image classification method based on multi-feature fusion, which uses a classifier to classify capsule endoscope images by fusing image color features and texture features. Compared with the state-of-the-art, this method adopts texture features combined with color information. The detection of invalid areas is an adaptive detection algorithm based on superpixels, which can automatically detect dark and bright areas in the image, realize feature extraction in effective areas, and avoid the interference of invalid information. Feature fusion adopts discriminative ability analysis technology to make the fused new features more discriminative. The invention can automatically classify the images of the capsule endoscope, thereby shortening the review time and reducing the burden on doctors.

Description

technical field [0001] The invention belongs to the technical field of image processing, and in particular relates to a capsule endoscope image classification method based on multi-feature fusion. Background technique [0002] Conventional hand-held endoscopy cannot involve the entire digestive tract, gastroscopy can examine the upper gastrointestinal tract, and colonoscopy can only examine the colon and rectum, so there are still parts of the intestinal tract that cannot be explored by conventional endoscopy. The emergence of capsule endoscopy allows the entire digestive tract of the human body to be observed. Usually, the patient swallows the capsule endoscope from the mouth, and through the peristalsis of the gastrointestinal tract, the capsule endoscope stays in the body for about 8 hours and takes images at a rate of 2 frames per second. The patient will carry a receiving device with him, and the images taken by the capsule endoscope will be transmitted to the receiver...

AI Technical Summary

Problems solved by technology

[0003] In order to overcome the deficiency that the existing technology cannot effectively realize the capsule endoscope image, the present invention provides a capsule endoscope image classification method based on multi-feature fusion that can effectively realize the capsule endoscope image. By extracting image features, the classifier is trained , Divide the capsule endoscopic image into several main parts such as esophagus, cardia, stomach, pylorus and duodenum, so as to reduce the burden of doctors' review and improve efficiency

Images