Design And Implementation Of Transfer Function For Volume Rendering

Scientific visualization is a new research field emerged in recent years. Volume rendering is a developed techniques in the visualization of three-dimensional data sets,which is widely applied in Geology, Meteorology, and Medical Imaging and other academic field. Transfer function whose role is to transfer the original values of the data set to optical properties, such as color and opacity, is an essential step in volume rendering, and directly affects the quality of volume rendered images. However, because of the complexity of the volume data and the diversity of transfer function, finding a universal and intuitive transfer function remains very challenging.Traditional design of transfer function for segmented data sets is only based on the scalar values. This method, however, cannot selectively visualize the objects of interest, and is inconvenient for distinguishing multiple objects of interest with user's specific needs. To address this problem, simplified LH histogram is applied to design the transfer function for segmented data sets in this paper. This method breaks through the limitations of traditional way of transfer function for segmented data. Users can effectively and efficiently locate and select the regions of interest. But in the LH histogram, the different tissues in the boundaries cannot be distinguished clearly. The f-LH histogram is further proposed to solve the problem. And the efficiency of the proposed transfer function design methods is verified using the segmented human cardiac MRI data set.Because the ordinary volume data sets suffer from the effect of noise, the slice data are first preprocessed with non-linear filters to improve the quality of 3D volume data to facilitate the setup of the transfer function. Spatial information is also one of the most important attributes of the feature in the volume data. So the LH histogram is enriched with the spatial information to produce the spatial information enhanced LH histogram. Region growing method in this paper is used to setup the transfer function more semi-automatically. As the volume rendering algorithms usually suffer from large computations, evaluating the performance of the transfer function is a time-consuming process. In this paper, GPU-based ray casting algorithm is implemented, which will greatly improve the rendering speed, and the real-time feedback of the transfer function can be effectively visualized.Since the diversity and complexity of the design for the transfer function, it is required to combine the manual method and automatic method to setup the transfer function semi-automatically. Users are provided with the interaction interface and tools to select the regions of interest. The whole transfer function can be setup semi-automatically with their needs for real-time interaction with the rendering results.