Designing effective transfer functions for volume rendering fromphotographic volumes

摘要

Photographic volumes present a unique, interesting challenge fornvolume rendering. In photographic volumes, the voxel color isnpre-determined, making color selection through transfer functionsnunnecessary. However, photographic data does not contain a clear mappingnfrom the multi-valued color values to a scalar density or opacity,nmaking projection and compositing much more difficult than withntraditional volumes. Moreover, because of the nonlinear nature of colornspaces, there is no meaningful norm for the multi-valued voxels. Thus,nthe individual color channels of photographic data must be treated asnincomparable data tuples rather than as vector values. Traditionalndifferential geometric tools, such as intensity gradients, density andnLaplacians, are distorted by the nonlinear non-orthonormal color spacesnthat are the domain of the voxel values. We have developed differentntechniques for managing these issues while directly rendering volumesnfrom photographic data. We present and justify the normalization ofncolor values by mapping RGB values to the CIE L*u*v* color space. Wenexplore and compare different opacity transfer functions that mapnthree-channel color values to opacity. We apply these many-to-onenmappings to the original RGB values as well as to the voxels afternconversion to L*u*v* space. Direct rendering using transfer functionsnallows us to explore photographic volumes without having to commit to anna-priori segmentation that might mask fine variations of interest. Wenempirically compare the combined effects of each of the two color spacesnwith our opacity transfer functions using source data from the VisiblenHuman project