Study On The Technology And Algorithm Related To The Real-time Construction Of Wide Field View Of Medical Ultrasonic Image

In the field of medical imaging, medical ultrasonic imaging has exerted greatinfluence in clinic medical science on account of the advantages of real-time, non-invasive,nondestructive, cheap, simple operation, and high sensitivity, so clinicians take it as animportant base for clinical diagnose. However, owing to the low resolution of theultrasonic probe and the corresponding limitation on field of view, the observation resultsfor large area are not that satisfied, which force people to seek for a better modification toextend the field of view. Although hardware technology can be applied to expand the fieldof view, its practicability in ordinary medical ultrasonic imaging equipment is quite lowbecause of the expensive equipment cost.It is not only has theoretical significance, but also has the value of clinical applicationto study how to use software method to realize real-time ultrasonic image with large fieldview.This paper aims to study the theory and implementation methods of the systemmentioned above, that is to gain real-time image with large field view through matchingand fusing of the corresponding series of common real-time images with small field view.And then, we can present the whole organ or tissue in the same image and allow thedoctors to diagnose quickly and accurately. Therefore, the author of this paper, in theMicrosoft VC++6.0environment, using the MFC programming method studied severalcommon used noise filtering methods, and then presented and realized a new speckledenoising method based on The improved median value fast tabulation in3dimension.We focused on the filtering effect and running time compared to the other commonly usedalgorithms.Secondly, in the Microsoft VC++6.0+OpenCV environment，we studied theregistration and fusion algorithm for medical ultrasound image, Including theoreticalstudy and simulation experiment. A new registration method presented, which used theimproved1/2nearest neighbor search method based on SURF to obtain the imagematching points and improved RANSAC method was used to establish the registrationmodel. The method presented in this paper was compared with the commonly used Harris,and SIFT method. This new method not only reduced the image noise, but also weakened the dependence on map matching. Hence, the amount of calculation greatly decreased,accompanied by the improved robustness of noise, rotation, scale and luminosity andenhancement of registration accuracy.Thirdly, A new fusion method presented based on the mean value coordinates(MVC) modified Poisson equations solving method used to medical ultrasonic image fastfusion. This method presented in this paper was simulated and compared with thecommonly used methods in the environment of Matlab7.1The experiment results showed that the method proposed in this paper significantlyimproved the image matching accuracy, together with the enhancement of the matchingspeed compared with the common methods.