Volume Rendering On Boundary Surfaces Of3D Images

Volume rendering is a scientific visualizationtechnique based on opticalemission and absorption.At present, ray-casting algorithmprovides thehighest rendering quality possible, but gives poor real-time interactivitydue to its high computational complexity. One Ray-cast rendering algo-rithm based on GPU is realized in this paper taking GPU’s advantage infloat-point computation and hardware parallelization. Furthermore, im-provements has been made in this paper, compared with traditional light-ing models, using half-angle formula based on Blinn-Phong model in illu-mination calculations.With the increasing request to distinguish different targets in volume data,the design of the transfer function(TF) has been one of the hot spots aswell as the most difficult field of research. TF establishes the mapping fromthe feature information of scalar value to optical properties. Moreover,many researchers and practitioners of visualization technology pay moreattention to the display of the boundary surfaces of volume data. However,some structures cannot be extracted by selecting different regions in fea-ture space and assigning different optical properties to them, because the-se structures are connected spatially which means their feature infor-mation is also connected. An algorithm based on Marching-cube boundarysurface zero-cross extraction is proposed in this paper to re-build the spa-tial feature space hence assists the inter-actor to extract the boundary sur-face intuitively. In addition,another algorithm based on the visual model toimprove the extraction of the boundary surface, is employed to optimizethe design of the transfer function by involving the spatial information. Ex- periments prove that both algorithms serve as indications for inexperi-enced operators to select targets of interest from feature space.The boundaries of volume data with familiar attributes fall into two catego-ries: adjacent and disjunctive structures. It’s difficult to classify the sam-pling points of boundary owing to the PVE. This paper presented an algo-rithm that disconnects the adjacent boundary with connection domain anddilation operation. It is proved by the experiments that the abovemen-tioned theories could perform good separation of spatially adjacent objectsof the same feature.