Based On Improved Image Stitching Algorithm SURF Medical X-ray

It is important and essential for the doctors to obtain a complete anatomical image in their clinical diagnosis and treatment, which can be applied to the spine disease or other disease related to bone. As far as the traditional film image is concerned in current image system, it has not only a limited precision, but also a long time-consuming combination process. Therefore, it is fundamentally necessary to attain the partial X-ray image, which can be further integrated into a complete clear image. There was a breakthrough development in the digital imaging technology and biomedical engineering technology, which provided a strong support for the automatic and gapless combination concerning medical X-ray image. Furthermore, medical image combination not only assisted doctors in diagnosing and treating, but also had its practical value in the supervision of the disease change and the assessment of the treating plan.The main purpose of this thesis is to apply the improved Speed-up Robust Features (SURF), SURF to the medical X-ray image combination. First of all, the brief review is given about the researching results associated with the present situation, especially the symptom detection and the precise algorithm. And then there was a brief introduction about the principle and composing elements of the medical image combination as well as the classification of the geometrical change. Next, through analyzing the characteristics of the medical X-ray images, such as the non-uniformed brightness, easily-covered details and the low comparability, this thesis employed canny detection to improve the original SURF, which was applied to the medical X-ray image combination. However, all these were done based on a first processing of the low-quality images so as to highlight the details. At last, when the image combination had been completed, the gap left over was removed by the balancing fusion through fading in and out. Only through comparing the distance and deviation of the original images’ characteristics and the combined ones in the thirty experiments, could this algorithm be assessed objectively. And it had been proved to be more effective through the improved algorithm.