Algorithm Research And Software Implementation Of Capsule Endoscopy Video Reduction

Capsule endoscopy is a new type of medical testing equipment,which can be used to check the digestive tract of the human body.Compared with the traditional testing equipment,capsule endoscopy not only successfully removes the blind spot of human small intestine detection,but also examines the human digestive system painlessly and non-in'vasively,its appearance is of great significance for improving human digestive tract detection ability.In the process of digestive tract examination,capsule endoscopy takes tens of thousands of endoscopic images,from which doctors will spend a lot of time finding out the images that contain lesions information.This is a very heavy and tedious job for doctors.It is found that the continuous frames in the capsule endoscope video have high similarity,and there is a lot of redundant information between adjacent frames.Therefore,the method of redundancy screening can be used to reduce the size of capsule endoscope video so as to shorten the time of doctor's reading and improve the efficiency of diagnosis.Aiming at the reduction of capsule endoscopy video,the existing screening methods have the problems of low accuracy and slow processing speed.Low screening accuracy will lead to the fact that the video contains a small amount of redundant information or the key information frame is deleted which will increase the risk of misdiagnosis of doctors.Slow processing speed will lead to the screening speed of computer is slower than the doctor's reading speed,which is inconvenient to practical use.Through the characteristics analysis of capsule endoscopy images and summarizing the existing research results,this paper proposes a new video reduction algorithm,which can quickly and accurately screen redundant images from the capsule endoscope video stream.On the basis of this algorithm,a capsule endoscopy video reduction software is developed,which can reduce the video quickly and effectively.This research has a high theoretical and practical value.The research work accomplished and the main conclusions are as following:(1)This paper puts forward a new algorithm for image reduction of capsule endoscopy based on color and texture features.First,the images are converted into HSI space,and the similarity of the feature is measured by cosine angle for color matching;Then,the image pyramid is constructed,from which multi-scale texture features are obtained by use of improved local binary pattern operator;Similarity measurement of the texture is conducted by computing weighted Manhattan distance of the texture feature between two successive frames and calculating the proportion of area with different texture.Finally,a cascade classifier is used to complete the classification of the images.(2)Through the analysis of the classical local binary pattern operator and the characteristics of capsule endoscopy images,an improved local binary pattern operator LBP_max is proposed in this paper.The operator can effectively extract the local texture of the image,and its code value can reflect the gradient direction of the texture obviously,the operator has the characteristics of fast texture extraction,low feature dimension and strong robustness,which means that the operator is suitable for texture analysis of capsule endoscopy images.(3)In order to verify the accuracy of the algorithm and the ability of video reduction,the performance of the algorithm was tested using 2600 frames capsule endoscopy images.The test results show that the recall,precision and F-measure of this proposed algorithm reach 91%,87%and 89%respectively.The compression ratio of the video is about 44%,and the average processing speed is 0.055 seconds per frame.Compared with the existing algorithms,the proposed algorithm has a good performance of identifying the redundant images in the capsule endoscopy video.(4)Capsule endoscopy video reduction software is developed on Microsoft Visual Studio 2010 software development platform with open source visual library OpenCV provided by Microsoft.The software has a very good performance in video reduction;the software interface can display the processing image,a variety of process parameters and the final calculation results.