Stitching Algorithm Of Feature Based On X-Ray Image

In recent years, x-ray stitching technology has important theoretical significance and practical clinical value in the medical field. When doctors diagnose and treat patients’ scoliosis and lower limb in the artificial joint replacement. They need to get patient complete lower limb x-ray image and spine x-ray image. The large size of x-ray image information can help doctors to obtain a more comprehensive understanding of the lesion and the region of interest. However, due to the limitation of the x-ray equipment, only shooting to the large size of the x-ray image can meet the requirements. Then application of image stitching technology makes several small-sizes images synthesize large-size images. Finally, doctors can get integrated and seamless images so that they can better observe the location of the lesion and more effective treatment for patients.An x-ray image stitching algorithm based on feature points is presented in this paper, which solves the splicing problem of the x-ray images when noise and intensity inhomogeneity and small overlap area exist. First of all, the algorithm of image stitching each step is summarized and analysed, and the algorithms of image registration are studied and classified in this paper. Then, traditional Harris corner detection algorithm and improved Harris corner detection algorithm of x-ray image feature extraction based on automatic stitching method are presented. The algorithm of improved Harris can effectively remove boundary effect and the algorithm is faster of running time. But it is sensitive to noise. Therefore, the SIFT (Scale Invariant Feature Transform) operator is proposed. It is a strong stability operator, which can invariant to scale, translation and rotation and robust to noise, illumination and affine transformation. Through the experimental comparison, we can see the SIFT operator better meet the requirements. Using of normalized cross correlation algorithm achieving feature points coarse matching, and the introduction of robust random sample consensus (RANSAC) algorithm can remove false matching points and achieve feature points precise matching. Then, affine transformation estimate space transformation model is presented. The weighted fusion algorithm is applied to achieve smooth and seamless stitching.Following manual stitching method, semi-automatic stitching (manually marked points stitching) method and automatic stitching method are applied to achieve in the platform in this paper. Comparison of manual and semi-automatic stitching method, automatic stitching algorithm can reach a high success rate in the stitching of long bones, spine and lower extremity x-ray images.