Fast Volume Rendering and Deformation Algorithms

Volume rendering is a technique for simultaneous visualization of surfaces and inner structures of objects. However, the huge number of volume primitives (voxels) in a volume, leads to high computational cost. In this dissertation I developed two algorithms for the acceleration of volume rendering and volume deformation. The first algorithm accelerates the ray casting of volume. Previous ray casting acceleration techniques like space-leaping and early-ray-termination are only efficient when most voxels in a volume are either opaque or transparent. When many voxels are semi-transparent, the rendering time will increase considerably. Our new algorithm improves the performance of ray casting of semi-transparently mapped volumes by exploiting the opacity coherency in object space, leading to a speedup factor between 1.90 and 3.49 in rendering semi-transparent volumes. The acceleration is realized with the help of pre-computed coherency distances. We developed an efficient algorithm to encode the coherency information, which requires less than 12 seconds for data sets with about 8 million voxels. The second algorithm is for volume deformation. Unlike the traditional methods, our method incorporates the two stages of volume deformation, i.e. deformation and rendering, into a unified process. Instead to deform each voxel to generate an intermediate deformed volume, the algorithm follows inversely deformed rays to generate the desired deformation. The calculations and memory for generating the intermediate volume are thus saved. The deformation continuity is achieved by adaptive ray division which matches the amplitude of local deformation. We proposed approaches for shading and opacit adjustment which guarantee the visual plausibility of deformation results. We achieve an additional deformation speedup factor of 2.34~6.58 by incorporating early-ray-termination, space-leaping and the coherency acceleration technique in the new deformation algorithm