Perception based volume rendering for shape enhancement
Arie E. Kaufman, Dimitris Samaras, Lei Wang · 2013
Volume rendering has been used in various applications for visualizing 3D volumetric data. It can visually separate objects and reveal the shape of the objects in a 2D rendered image. The work presented herein enhances the volume rendering using perception-based methods. It separates the objects in the volume data based on their shape and visually separates them in the rendered image by encoding the importance of an object using the visual saliency of its color. It then conveys the local and global shape of an object, as well as depth perception using specialized lighting. The shape information of a volumetric dataset is captured using the proposed cumulative heat diffusion (CHD) that is evaluated by a volume gradient operator (VGO). The VGO captures the local shape information while the global shape is obtained through the cumulative heat diffusion. The CHD simultaneously provides the shape information in different scales, from local to global, for each voxel. Furthermore, a Monte-Carlo based method is proposed to accelerate the capture of the global shape information. The shape information in multiple scales constructs a multi-dimensional attribute space. To visualize the multi-dimensional attribute space for separating the objects based on the shape information, a method is presented to display the relationship of objects in attribute space using a 2D modified dendrogram (MD), with sub-trees to represent clusters. It allows the user to design a transfer function in an intuitive and informative manner in 2D, instead of working in the multi-dimensional space, where the relationship of the objects is difficult to ascertain. Moreover, the granularity of the MD is controlled by the user, allowing him to efficiently create a transfer function in a coarse-grained MD and to fine tune it with finer-grained MDs for object separation. Then, the objects are visually separated using an automatic color design approach. The color design approach makes the important object appear to be visually more salient in the rendered image. The color harmony of the rendered image is simultaneously preserved. Finally, the shape of an object is conveyed by our lighting system, which contains a key light and several accessory lights. The key light conveys the visual cues for the perception of local and global shape, as well as depth. The accessory lights increase the perceptibility and enhance the visual cues for local shape perception. Furthermore, the accessory lights are generalized to per-voxel accessory lights (two accessory lights for each voxel). The per-voxel accessory lights make it possible to control the revealed local shapes based on the importance value of each voxel. Meanwhile, the key lights are upgraded to curved lights, which traverse the volume data along curved paths, to simultaneously enhance the visual cues for the perception of the global shapes in different scales. The global impression is preserved during the configuration of the per-voxel accessory lights and the curved lights.