Research And Implementation Of Large-scale Data Volume Rendering And Multi-dimensional Transfer Function Technique

Volume rendering is one of the main methods in scientific visualization field. The intrinsic information of volume dataset can be visualized interactively by volume rendering. Volume rendering is also known as direct volume rendering, for no intermediate geometry primitives are generated during its visualization process. It can generate the image of dataset with high quality details and context. Volume rendering can be applied in many fields including medical science, meteorology and geological exploration.Transfer Function design is one of the most important procedures in volume rendering. Transfer function maps data values to display attributes, such as color and opacity. Different materials in dataset can be separated and displayed by a good transfer function. But traditional way of designing transfer function is trail and error, which is complicate and not intuitive. In this thesis, region-grow based multi-dimensional transfer function design method has been proposed. This method can greatly simplify the design task and lead to a good transfer function. Another problem of transfer function is that spatial information is not fully utilized. So we adopt a spatial information based multi-dimensional transfer function generation algorithm, which utilizes information both from spatial domain and transfer function domain to classify material automatically.As the development of scientific instruments, the size of dataset generated has been increased dramatically. Such large scale datasets pose immense challenges to volume rendering. In this paper, we use multi-resolution techniques for volume rendering, which manage original dataset in hierarchical tree structures. This method not only decreases storage limitation of volume rendering, but also increases rendering speed without sacrificing rendering quality. We improve selection rules and rendering rules in multi-resolution volume rendering process. A multi-dimensional transfer function based cull rule has also been proposed to further increase rendering speed.Finally, a general purpose and extensible volume rendering framework has been carefully designed and developed. Based on this platform, we implement several transfer function design algorithms and the multi-resolution volume rendering algorithm, which utilize GPU-based hardware acceleration method. The framework shows high rendering speed and image quality after testing by many datasets.