Research On Three-Dimensional Imaging Based On Volume Rendering

3D visualization technology of medical images is the process that 3D images of tissues and organs in a person's body have been reconstructed by computer which uses human visual medical characteristics, according to the 2D medical image slices. At present it is widely used in diagnosis surgery planning and plastic simulating and artificial limb surgery, radiotherapy planning, and teaching in anatomy. The study of 3D reconstruction from medical images has important significance on science and value in practical application.3D reconstruction from medical images can generally be classified into surface rendering and volume rendering. Compared with surface rendering in 3D reconstruction, volume rendering techniques can concentrate on visualizing internal features of choice and surfaces at the same time. So it has drawn extensive attension in recent years and has become one of the most researched topics.Volume rendering is still in his infancy, and many problems are yet to be solved. This paper addresses research on image formation based on volume rendering algorithm. The main contributions are summarized as follows:Firstly, this paper introduces the research background and covers the most up to date information while also including the developing trends of 3D reconstruction.On the basis of introducing classification of volume rendering, this paper narrates some algorithms, including ray-casting, splatting, shear-warp, 3D texture-mapping hardware, fourier volume rendering, hartley volume rendering, fourier-wavelet volume rendering, ray tracing in wavelet space, wavelet-based splatting etc, and then evaluates the characteristics and performance of all volume rendering algorithms for 3D reconstruction of medical image in a comparison table. Then, the author find out some problems of image formation based on volume rendering algorithm.In the visualization process of a three-dimensional dataset, the better view can improve both the speed and efficiency of the data understanding. The ray-casting algorithm is an important algorithm of volume rendering. Due to its maturity and the high quality of image, it has already been applied in many medical visualization process systems. The author proposes a view selection method for volume rendering algorithm. It combines the formulation of entropy with the ray-casting algorithm to objectively evaluate the quality of view. This method takes account of the transfer function, the data distribution and the visibility of the voxels.The results from the experiment demonstrated that it can not only be used to pick up the most informative view for a given scene when the visualization process is non-interactive, but it can also enable the user to locate the better viewpoint in order to investigate further when the visualization process is interactive.Frequency domain volume rendering include: fourier volume rendering, hartely volume rendering, fourier-wavelet volume rendering. The resulting images of frequency domain rendering are easy to appear aliasing. This aliasing occurs in the spatial domain as a simple overlapping of copies of the original dataset. So this paper proposes: the volume data preprocessing, re-sampling and the choice of reconstruction filter need to be studied and combined to eliminate the ghosting. The experimental results demonstrated that this method is effective and the ghosting disappears, within a certain scope of the screen.This paper introduces the fourier-wavelet volume rendering algorithm in detail, and partially realizes it. The resulting images of fourier-wavelet volume rendering are x-ray like images and lack in depth information. This paper proposes a negative index depth cueing approach implemented to compensate this problem. The experimental results demonstrate that it is effective and enhances the visual quality. At the same time, the major shortcoming in images quality of the fourier-wavelet volume rendering is self-occlusion. View selection method is suited to this case and pick up the most informative view. The experimental results demonstrated that this method can reduce self-occulsion and allow the user to find a good viewpoint, then decide to arrive ('progressive refinement'). Especially to the client-sever visualization system, the server implements the view selection method to find out a good viewpoint, then interacts with clients which can improve interaction efficiency.