Capsule endoscopic image retrieval based on wavelet transform
Side Liu · Modern Digestion & Intervention · 2011
Objective To explore the effectiveness of content based capsule endoscopic image retrieval of suspected cases.Methods The endoscopic image texture feature was extracted using wavelet transform and some relevant improvements were made.The images from each channel were processed in RGB and HSI spaces using three layers of wavelet decomposition,and low-frequency information of each channel signals was extracted,assigning different weights to each component.Each channel's low-frequency information combined image colors with textures information,and took it as a divisional endoscopic image feature vector,then,similarity was depicted between inspected images and standard images by classic Euclidean distance.Results The experimental results showed that tHSI paper method had better results in retrieving endoscopic image by comparing three methods'results.Conclusions Aiming at capsule endoscopic image characteristics,tHSI paper could fairly reflect the spatial distribution information of capsule endoscopic image color and accorded with doctors' vision.