Study On Segmentation Of X-ray Image Based On SVM

X-ray images in Orthopedics are the main basis for the diagnosis of bone, but X-ray images have been reading by doctors. Because of subjective reasons the result of judging is wrong. With the development of digital image processing technology and computer technology, computer aided interpretation of X-ray images is proposed which improves diagnostic efficiency and accuracy to some degree. It is the basis of computer aided interpretation to segment skeletal tissues through muscle tissues. This paper studies how to correctly extract the bone tissues from X-ray images.Because of uneven regional gray and low contrast in X-ray images, it is difficult to segment bone tissues from X-ray images. This paper proposes two segmentation methods of X-ray images, one based on gradient and the other based on SVM.In the method based on gradient of X-ray images, we only deal with the border pixels according to target border pixels having larger gradient. Look for the largest gradient point by searching the gradient image as the starting point and the number of targets depend on the tracking numbers. After one target is segmented, the eight pixels of its neighborhood of the target boundary pixels are dealt with. After three tracings, we realize segmentation of X-ray image. But this method is sensitive to initial point and direction of tracking.In the method based on SVM, image pre-processing method including image enhancement and objectives positioning is designed which is suitable for classification for support vector machine. Sample selection, feature extraction, kernel function and parameters are discussed, also the error rate are calculated in different cases. Finally, bone issues are correctly segmented through automatic selection of training samples which is through C-means clustering method and nine-dimensional feature vectors which are neighborhood gray values.