A Study Of 3D US Image Data Reconstruction And Image Recognition

3D ultrasound imaging of human body anatomy is more reality for doctors, and it can provide a virtual endoscopic function and determine more precise location of the interested object to assist the diagnose of doctors. Moreover, it has many other available superiority than the conventional brightness model imaging system. Thus, the research on the reconstruction method in 3D imaging is significant for the application in practice. With the more and more widely application of imaging in medical, to read and understand these images, it requires the doctors have special knowledge in anatomy, electronic technique, imaging technique and acoustic area. Now, the processing and analysis of these medical images becomes an important problem.This dissertation aims at the reconstruction method based on rotation scan, a scanning mode requires the highest calibration accuracy, and a reconstruction model with parameters and software calibration was developed. This method degrades the requirement of hardware calibration accuracy, and can obtain a high accuracy 3D image with low calibration and manufacture level. In addition, some primary research work has been done in image target recognition in this dissertation. The main contents of this thesis include:(1) It reviewed the reconstruction systems with different scan methods and evaluated them.(2) Analyzed the reason producing reconstruction errors, and established the mapping function from the image with errors to the image without error.(3) Established a model of 3D image reconstruction with the parameters and software calibration based on rotation scan, and defined an image evaluation function, and resolved the value of those parameters using optimization algorithm. Then completed the high accuracy 3D image reconstruction with low calibration accuracy and manufacture level.(4) Introduced the image recognition and classification, and proposed a fast algorithm of a new contour descriptor. Experiment demonstrated that the new descriptor is robust to contour noise and has the invariability of translation, rotation and scale variation.(5) Pointed out the future work that should be done in the future.