| Epigastralgia(stomach ache) is a disease occurs near the pit of the stomach,which is frequently in clinical. Endoscopy is the most common and accurate method for stomach diagnosis. Epigastric pain is polymorphic under the endoscopy. Therefore, this paper will analyze epigastralgia features data based on endoscopy image, so as to realize the computer aided diagnosis.The change of gastric pathological has a great degree of relevance between epigastralgia course, therefore gastric texture is the main identification features of epigastric pain. Whether it is based on the statistical approach or model-based method, or spectrum, the classic image feature extraction algorithm prototype can not expression patterns and characteristics of the original input image reasonably and effectively. This paper study the image processing method based on the regional characteristics,which is applied to the endoscopy image, in view of Haar-type features can count local features simply and effectively.This paper will introduce it into local binary pattern(LBP) to enhance the performance of texture feature extraction.After extracted image feature, the support vector machine act as the classifier identifier. Extracting the histogram of image texture through the Haar type LBP algorithm,which act as a feature vector of SVM to make a final determination. Experiments show that to solve this problem with less endoscopy image sample and fewer features can be extracted, endoscopic image after HLBP texture feature extraction algorithm has better categorical for SVM.By the way, in order to improve the accuracy of classification, taking four specific locations into account,such as the tongue, lesser curvature, angle notch, pylorus, and using of vote principle to make a comprehensive judgment. Experimental results show that, compared to traditional LBP texture analysis,the identify of epigastric pain symptoms based on HLBP image processing method can keep balance between efficiency and accuracy with rate of 95.8%. |
PDF Full Text Request