Report

Shape-aware Volume Illustration Wei Chen, Zhejiang University, Purdue University Aidong Lu, University of North Carolina at Charlotte David S. Ebert, Purdue University Contents Motivation & Related Work The Main Idea Algorithm & Implementation Experimental Results Conclusions & Future Work Motivation Illustration is a visual representation Motivation Previous work on simulating appearance Rendering primitives: point, line, surface... drawings [LEM 02, BKR 05, FBS05] Motivation Previous work on simulating appearance Visibility guided or selective illustration: [DWE03, CSC06, VKG04, BG06, BGKG06, ONOI04] Motivation Previous work on simulating appearance System work: VolumeShop [BG05], IVIS[SDS05]… Motivation Most previous work relies on transfer functions to express important features Expressiveness is greatly influenced by the shape and shape variations Related work Volume Illustration Incorporate NPR techniques into volume rendering. Ebert et al. [ER00] Illustrate the internal structures by synthesizing a 2D texture on the cutting planes of polygonal models. Owada et al. [ONOI04] Hardware-accelerated volume illustration [SE03,HBH03, BG05, SES05, BG06] Related work Volume Illustration Related work Shape Representation and Processing Boundary representation Volumetric Representation Voxelization /Distance field computation Marching Cube Related work Volumetric Manipulation Traditional GPU-based volume deformation techniques [RSSSG01], [WRS01] Feature-aligned volume deformation [CSC06] VolumeShop system [BG05] Related work Example-based Modeling and Rendering Texture Synthesis [WM01] Image & Curve Analogy [HJO01] [HOCS02] Mesh Contour Analogy [ZG04] Example-based Volume Illustration [Lu2005] The Key Idea Creating volume illustration in a shapeaware manner A shape-aware volume representation Curve analogy based shape deformation Shape-aware volume illustration The pipeline Volume Segmentation Available Segmentation Approaches Manual segmentation Threshold-based or TF guided segmentation Level-set based segmentation (ITK) Volume Binarization 0 The bone 2 1 The foot dataset The skin Volume Filtering Volume Filtering For Binary Volume Image Morphology Algorithms Level-set based segmentation (ITK) [Whi00] Geometric Processing Feature Preserving Mesh Processing Mesh smoothing [JDD03] Mesh repairing [Ju04] Mesh simplification [ZG02] #V 229,298 #V 13,689 Geometric Processing Another example for the hand model Curve Analogy Based Surface Deformation Shape Variations are to be considered Using gradient domain based surface deformation techniques [SLCo05,YZX04] Curve Analogy based shape deformation [ZHS05] Introducing a proxy surface and connecting both surfaces with mean value coordinates [JSW05] ? Curve Analogy Based Deformation For a model M0, generate a simplified model M1 Generate the mean value coordinates for each vertex of M0 associated with M1 Specify a curve C1 in M1 and project it to the 2D plane Draw a curve C0 in the 2D illustration Deform C1 with respect to C0 Deform M1 with the deformed C1 by the mesh deformation algorithm Deform M0 by applying the mean-value coordinates to the deformed M1 Simplification M0 (#V 11067) Deformation M1 (#V 1917) Deformed M1 MVC Deformed M0 Curve Analogy Based Deformation Apply deformation to another model Deformation Skin (#V 1917) Computing Mean Value Coordinates Bone (#V 13689) Deformed Skin Applying Mean Value Coordinates Deformed Bone (#V 13689) Curve Analogy Based Deformation …… Specify a curve C1 in M1 and project it to the 2D plane Draw a curve C0 in the 2D illustration Deform C1 with respect to C0 using Laplacian editing [SLCo05] Deform M1 with the deformed C1 by the mesh deformation algorithm [ZHS05] …… C1 Deformed C1 C1 C0 C1 C0 M1 Deformed M1 Shape-aware Volume Representation A combination of a distance volume and a segmentation volume Each voxel records a distance value and a segmentation identification The distance volume The segmentation volume Their composition Shape-aware Volume Representation Compute the signed distance volume Construct an auxiliary octree grid to accelerate the computation of the distance volume [Ju04] Encode the distance as an unsigned integer Integrate all polygonal models into a single model, and compute its distance volum. Compute a distance volume for each individual object Shape-aware Volume Representation Generate the segmentation volume Based on the computed distance volumes Each voxel of the segmentation volume is first initilized as zero For each distance volume of the ith model, check the sign of each voxel • If it is negative, the corresponding voxel in the segmentation volume is assigned an identification i 6 5 4 3 2 3 4 5 The input model i 5 4 3 2 1 2 3 4 4 3 2 1 0 1 2 3 3 2 1 1 2 3 4 5 2 1 0 0 1 2 3 4 1 0 -1 -1 0 1 2 3 0 -1 -2 -2 -1 0 1 2 -1 -2 -2 -2 -2 -1 0 1 0 -1 -1 -1 -1 -1 -1 0 1 0 0 0 0 0 0 1 2 1 1 1 1 1 1 2 6 5 4 3 2 1 2 3 The distance volume i The segmentation volume Offset Volume Generate an offset volume by choosing all points satisfying dist(p)<t Useful to illustrate the object boundary May build a thin offset volume form each distance volume Blue: muscle Yellow: bone Green: bone boundary Red: skin Benefits of New Representation Gives a novel explanation to the data and yields a direct expression of shape Reconstruct smooth boundaries by exploiting the information of the distance volume Be able to distinguish individual objects Shape-aware Volume Illustration The representation and deformation scheme can achieve two goals. Suitable for applying various rendering styles to different regions of interest. Mimic artistic styles for object boundaries. Any volume rendering system can render the new representation Our implementation is based on IVIS volume illustration system [SDS05]. The volume is encoded in two volume textures. Shape-aware Volume Illustration The Uniform Illustration Equation Solid Texturing Color Shading Opacity Modulation Directly interpolate the computed colors (in RGBA space) on the eight nearest voxel centers Yields better results Takes about eight times the computational cost Experimental Results P4 3.2 GHz, 1.5G RAM nVidia Quadro FX 3400 Cg shading language Data #input #Output #Segments Brain 78x110x64 128x128x128 8 Kidney 256x256x64 256x256x64 2 Bunny 512x512x361 256x256x256 2 Head 256x256x256 256x256x256 4 Hand 256x128x256 256x256x256 6 Experimental Results IVIS system 3D texture slicing number: 1000 Image resolution: 480x480 Data I Brain II III FPS 2.0 s 5.0 s 480 s 2.5 Kidney 1.2 s 2.0 s 110 s 12.0 Bunny 1.2 s 2.0 s 620 s 2.0 Head 2.0 s 3.0 s 750 s 3.0 Hand 2.0 s 5.0 s 680 s 6.0 Foot Deformation Shape-aware Volume Rendering Bunny Deformation Hand Example MRI Brain Data Example Kidney Data Example Conclusions Contributions 1 The idea Incorporating shape into volume illustration 2 Algorithm •New representation •Curve analogy based deformation •Mixed volume illumination equation 3 Applications •Enhance the creation capability •Allows efficient illustration of the deformation of solid space Future Work More efficient algorithm Represent and learn intrinsic artistic shape styles from hand-drawn images Optimize computing of the distance volume Deformation-driven volume illustration of dynamic scenes Model-based volume illustration for special objects Acknowledgements Tao Ju (Washington University in St. Louis) Kun Zhou (Microsoft Research Asia) Xinguo Liu, Jing Huang (Zhejiang University) Nikolai Svakhine (Adobe) Oliver Deussen (Uni-Kanstanz university, Germany) Stefan Bruckner (Tu Wien, Austria) The Hand dataset is the courtesy of Tiani Medgraph, Austria. NSF of China (No.60503056) DOE DE-FG02-06ER25733, NSF 0633150, EPA VisualGrid NSF Grants 0081581, 0121288, 0328984, and the U.S. Department of Homeland Security. [email protected] http://web.ics.purdue.edu/~chen23 Work Pipeline of Volume Illustration 3D/4D Data Field Raw Data Data Acquisition ? Input Data Data Preprocess ? Final Result Interactive Illumination ! 2D Illustration ? Post-process Work Pipeline of Volume Illustration 3D/4D Data Field Raw Data Data Acquisition Input Data Data Preprocess ! Final Result Interactive Illumination 2D Illustration Post-process