||Image processing and analysis algorithms are widely used in medical systems to analyze medical images to help diagnose the disease of the patient. This thesis covers one of the demanding problem of medical systems: Stitching of X-ray Images. The flat panel of X-ray system can not cover all part of a body, so image stitching is incorporated in the medical system to combine two or more X-ray images and get a single high resolution image of the body part. The output of this thesis work is to develop a real time user interactive stitching application which works for X-ray images of different intensity and orientation.
The stitching process consists of two main steps: Image Registration and Blending. Many existing image registration methods are based on classical pixel-wise matching(exhaustive matching) or Chamfer matching which are slower registration methods. So, key-points based methods(Harris, SIFT,SURF) are selected to make the registration faster. The features of key-points computed from the key-point extractors are then used for matching.The exhaustive nearest neighborhood(NN) based matching method has been evaluated with approximate nearest neighborhood method(ANN). The further simple tests(Ratio Test and Symmetry Test) are employed to improve the accuracy of the matches for better registration result.
The overlapping area of the registered images are blended to remove the seams and discontinuities of the composite image. The thesis evaluates two blending methods: Pyramid Blending and Alpha Blending in terms of accuracy and computational complexity. The advanced blending with blending masks for complexly aligned images is giving a very encouraging result.