节点文献
基于网络的医学CT三维重建系统关键技术研究
【作者】 虢伟；
【导师】 章勤；
【作者基本信息】 华中科技大学 ， 计算机应用， 2007， 硕士
【摘要】 医学图像三维重建是目前医学可视化领域的研究热点,属于多学科交叉,涉及到计算机图形学、图像处理、生物医学工程等多种技术,在诊断医学、手术规划及模拟仿真等方面有广泛应用。基于网络的医学CT三维重建系统,通过将CT三维重建技术,基于图像渲染的技术以及网络通信技术融合,为目前医疗诊断可视化提供了一种新的手段。系统采用高速服务器进行大数据量的医学CT三维重建处理、数据传输以及处理结果的实时渲染,系统的客户端结合服务器端内存渲染的结果与基于图像的渲染技术,让用户实时查看服务器端三维重建的结果,服务器端还提供给用户处理结果下载的功能。在三维重建技术实现方面,研究改进了移动立方体算法,新增了8种基本拓扑架构,消除了传统移动立方体算法法线的二义性。另外,针对三维重建后可能出现的后期处理大规模数据,对传统计算传递闭包的分组算法作了改进,将算法复杂度由O(n3)降低到O(kn2)。在客户端的渲染实现方面,采用了一种的新的基于多幅深度图像的渲染算法,不仅解决了基于单幅深度图像渲染需要不断更换参考帧的弱点,而且还成功的解决了由于网络的不稳性而不能远程实时浏览三维重建结果的难题。通过对系统的服务器的功能测试与算法的性能测试,新改进的CT三维重建算法消除了原有算法所带来的黑斑,对512×512×34的CT序列重建只需要2s左右的时间。系统的客户端基于多幅深度图像渲染算法,在普通pc机上帧率可达到20fps以上,而且渲染的结果十分接近几何真实渲染的效果。
【Abstract】 3D reconstruction for medical images is a hot subject of medical images processing, belonging to multi-disciplinary subject, involved in computer graphics and image processing in biomedicine engineering. It is widely used in simulation of surgery, diagnoses, emulator and etc.The 3D reconstruction,image-based rendering and network are combined by a new architecture based on internet. It turns on a new way for medical visualization. The high speed servers are used to deal with CT 3D reconstruction in this architecture. And the technique of image-based rendering and internet are joined in its client. Because of this technology , the user can remotely view the result of 3D reconstruction. In addition, it is convenience for user to download the result.The implement of the 3D reconstruction in server is based on the advanced marching cubes. There are another 8 base cases are add to the algorithm. It is very successful to solve the confusion of the face in the marching cubes. On the other side, we improve the algorithm of the segmentation. The rendering in client are implemented by a new depth images-based rendering. This algorithm avoids the weakness of the single image-based rendering. It is necessary to change the reference image for the single image-based rendering. And this algorithm is not affected by the network. It can solve the problem of the 3d remote view.After functional and performance tests in server, the black spots have been removed with the advanced algorithm. It takes only 10 second for 512×512×34 CT images. The images-based rendering can reach the speed of more then 20fps in client. The result of its rendering is very nearly to the drawing of geometry-based rendering.