The Research In X-ray Medical Image Stitching With Improved SIFT Algorithm

Image stitching technology can be a very good solution for the contradiction between the image vision and resolution in X-ray projection imaging because of the limitations of imaging equipment and other factors.It can provide clear,high measurement precision image in order to provide a reliable basis for clinical diagnosis,preoperative plan,postoperative curative effect evaluation and so on.This article mainly studies the digital X-ray medical image stitching problem,the main contents are as follows:1.Briefly introduce the basic process of image stitching.List the classics algorithms of image acquisition,preprocessing,registration and fusion in order to provide theoretical basis for image stitching.2.Deeply study the concept of scale space and the application of differential operator in feature detection.Puts forward a new differential operator,introduce the concept of characteristic scales and verify the stability of the new differential operator from the theory and practice.Compare the advantages and disadvantages of the new differential operator and the laplace operator through practical application.? The new differential operator is more sensitive to characteristics that detects more feature points under the same condition.? Laplace operator is more sensitive to edge response.As a result,the proportion of feature points out with the new differential operator is less than that with the Laplace operator.? The Laplace operator can reduce a certain computation time,because the SIFT algorithm directly subtract the two adjacent dimension image instead of doing laplace transform.? In practical application,the characteristics scale of the new differential operator is relatively smaller,the proportion of the noise points may be a few taller,the antinoise ability is weaker and the specific situation remains to be further research.3.Introduce the concept of characteristic scale,describe the feature points on the corresponding characteristic scale space in order to highlight the characteristics.4.Quantitatively describe the feature points with the method in SIFT which is invariance of translation,rotation and scaling.The structure of K-d tree is used in feature point matching in order to increase efficiency.According to the actual situation,put forward the simple algorithm which is based on the statistics to eliminate mismatch.After testing,the new algorithm is stable and feasible,and the operation time is shorter.5.Cutting the original images into small size for image stitching to get the transformation parameters according to mechanical precision.Stitch and fuse the original images by the transformation parameters with the relation between the original images and the small size images and the operation time is shorter.