NPR Today ArtBased Rendering of Fur Grass and
NPR Today • “Art-Based Rendering of Fur, Grass and Trees”, Michael A. Kowalski et. al. , SIGGRAPH 99 • “A Non-Photorealistic Lighting Model for Automatic Technical Illustration”, Amy Gooch, Bruce Gooch, Peter Shirley and Elaine Cohen, SIGGRAPH ’ 98 04/01/05 © 2005 University of Wisconsin
Art-Based Rendering of Fur, Grass and Trees Michael A. Kowalski et. al. Presented by Scott Finley 04/01/05 © 2005 University of Wisconsin
Introduction n n Highly detailed objects containing fur, grass etc. are expensive to render. This paper attempts to use a well known artistic technique to indicate complexity with simple shapes. 04/01/05 © 2005 University of Wisconsin
Goals 1. 2. 3. 04/01/05 Give the designer control over the style. Simplify modeling by making rendering strategy an aspect of modeling. Provide interframe coherence for the styles developed. © 2005 University of Wisconsin
Prior Work n n Reeves used particle systems to create complex geometry from simple shapes. Alvy Ray Smith used particles and “graftals” to create his “Cartoon Tree”. ¨ Badler n and Glassner generalized the idea of graftals. This paper uses a modified version of difference images proposed by Salisbury et al. 04/01/05 © 2005 University of Wisconsin
More Prior Work n Meier’s particle-based brush strokes showed non-geometric complexity and fixed particle spacing on objects. 04/01/05 © 2005 University of Wisconsin
Software Framework Models are broken into “patches”. Each is rendered by a procedural texture. Two types of “reference images” are used: n n Color reference image 2. ID reference image 1. ¨ ¨ 04/01/05 Provides patches with list of pixels Can be used to find visibility of known point © 2005 University of Wisconsin
Graftal Textures Place fur, leaves, grass etc. on geometric models. n Need to be drawn in a controlled way in screen space. n Need to stick to models for inter-frame coherence. n 04/01/05 © 2005 University of Wisconsin
Before 04/01/05 © 2005 University of Wisconsin
After 04/01/05 © 2005 University of Wisconsin
Difference Image Algorithm n Each patch draws its region in the color reference image. ¨ Darker areas indicate more “desire” for graftals to be placed. ¨ In the examples here we want graftals along the silhouettes. Render with point light at camera n Can be done explicitly by designer. (Bear’s feet) n 04/01/05 © 2005 University of Wisconsin
Difference Image Cont. n Graftals are placed according to the desire in the color reference image. ¨ This allows screen space density to be controlled. n Bin all the pixels according to the desire level and start placing graftals on the pixels with the highest desire. 04/01/05 © 2005 University of Wisconsin
Creating Inter-frame Coherence n Need to be sure that graftals persist across frames to avoid extreme noise etc. ¨ In first frame place graftals according to DIA. ¨ In further frames attempt place graftals from previous frame. ¨ Place new graftals where needed according to the DIA. 04/01/05 © 2005 University of Wisconsin
Subtracting Blurred Image n When a graftal is placed it subtracts a blurred “image” of itself from the reference image. ¨ Graftals are treated as a point for this. The “image” is a Gaussian dot. n Size of the dot is proportional to the screen space area of the graftal. 04/01/05 © 2005 University of Wisconsin
Graftal Sizing Graftals can be set to scale according to perspective, have a constant size, or somewhere between. n Graftal size can be reduced if it tries to draw itself but there isn’t enough desire. n 04/01/05 © 2005 University of Wisconsin
Drawing Graftals n n Fur graftals can be drawn at different details with triangle strips. Drawing happens in surface normal plane. ¨ Detail depends on angle to viewer 04/01/05 © 2005 University of Wisconsin
Future Work n Reduce flicker/popping as graftals enter and leave. Use alpha blending ¨ Put graftals on the back of objects ¨ Use several layers of statically placed graftals ¨ 04/01/05 © 2005 University of Wisconsin
New Styles n Dual layered fur ¨ Suggests complex lighting 04/01/05 © 2005 University of Wisconsin
04/01/05 © 2005 University of Wisconsin
A Non-Photorealistic Lighting Model for Automatic Technical Illustration Amy Gooch, Bruce Gooch, Peter Shirley, Elaine Cohen SIGGRAPH ’ 98 (presented by) Tom Brunet University of Wisconsin-Madison CS 779 04/01/05 © 2005 University of Wisconsin
Background • Various NPR Techniques – Cassidy J. Curtis, Sean E. Anderson, Kurt W. Fleischer, and David H. Salesin. Computer-Generated Watercolor. In SIGGRAPH 97 Conference Proceedings, August 1997. – … • Technical-like – Takafumi Saito and Tokiichiro Takahashi. Comprehensible Rendering of 3 D Shapes. In SIGGRAPH 90 Conference Proceedings, August 1990. – Doree Duncan Seligmann and Steven Feiner. Automated Generation of Intent-Based 3 D Illustrations. In SIGGRAPH 91 Conference Proceedings, July 1991. – Debra Dooley and Michael F. Cohen. Automatic Illustration of 3 D Geometric Models: Surfaces. IEEE Computer Graphics and Applications, 13(2): 307 -314, 1990. 04/01/05 © 2005 University of Wisconsin
Contributions • Reduction of dynamic range needed to portray shape • NPR method for appearance of metal 04/01/05 © 2005 University of Wisconsin
Diffuse Shading ka: ambient illumination kd: diffuse reflectance 04/01/05 © 2005 University of Wisconsin
Highlights and Edges 04/01/05 © 2005 University of Wisconsin
Diffuse w/ Edges/Highlights ka: ambient illumination kd: diffuse reflectance 04/01/05 © 2005 University of Wisconsin
Alter Shading Model • Want to keep lighting from above • Extend shading across entire sphere: • Finally, mix a coolwarm hue shift with a luminance shift 04/01/05 © 2005 University of Wisconsin
Near Constant Luminance 04/01/05 © 2005 University of Wisconsin
Color & Luminance Shift 04/01/05 © 2005 University of Wisconsin
Maintains ‘Color Name’ 04/01/05 © 2005 University of Wisconsin
Metal Appearance • Milling creates anisotropic reflection • Pick 20 strips of random intensity [0, . 5] • Linearly interpolate 04/01/05 © 2005 University of Wisconsin
Metallic, Anisotropic Reflection 04/01/05 © 2005 University of Wisconsin
Approximate in Open. GL • Two opposing directional lights: • (kwarm - kcool)/2 • (kcool - kwarm)/2 • Ambient: (kcool + kwarm)/2 04/01/05 © 2005 University of Wisconsin
Other Results/Questions 04/01/05 © 2005 University of Wisconsin
04/01/05 © 2005 University of Wisconsin
Computer Generated Watercolor (Siggraph 1997) Cassidy J. Curtis Sean E. Anderson Joshua E. Seims Kurt W. Fleischer David H. Salesin 04/01/05 © 2005 University of Wisconsin
Watercolor Effects • • • Drybrush Edge Darkening Backruns Granulation Flow Effects Glazing 04/01/05 © 2005 University of Wisconsin
Previous Work • David Small. Simulating watercolor by modeling diffusion, pigment, and paper fibers. February 1991. • Qinglian Guo and T. L. Kunii. Modeling the diffuse painting of sumie. In T. L. Kunii, editor, IFIP Modeling in Comnputer Graphics. 1991. • Julie Dorsey and Pat Hanrahan. Modeling and rendering of metallic patinas. 1996. 04/01/05 © 2005 University of Wisconsin
Improvements • More complex paper model • Better compositing (KM) • Three layer simulation – Shallow-water layer – Pigment disposition layer – Capillary layer • Painting represented as layers of wash (dried watercolor) 04/01/05 © 2005 University of Wisconsin
Algorithm Overview For each time step: • Move. Water – Update. Velocities – Relax. Divergence – Flow. Outward • Move. Pigment • Transfer. Pigment • Simulate. Capillary. Flow 04/01/05 © 2005 University of Wisconsin
Algorithm Update. Velocities • Height gradient used to modify velocities • Simulate shallow water flow using Euler Method and standard flow equations • Velocity of pixels outside wet area mask are set to zero Relax. Divergence • Distribute fluid to neighboring cells 04/01/05 © 2005 University of Wisconsin
Algorithm Flow. Outward • Remove water from each cell • p = p – n * (1 – M’) * M Move. Pigment • Pigment distributed to neighboring cells 04/01/05 © 2005 University of Wisconsin
Algorithm Transfer. Pigment • Pigment is deposited or lifted – Density of pigmentation – Staining power – Granulation Simulate. Capillary. Flow • Transfer water from shallow water layer to capillary layer • Water is diffused to neighbors in the capillary layer • Wet area mask updated 04/01/05 © 2005 University of Wisconsin
Rendering • Layers combined using Kubelka-Munk method • Interactive pigment creation system • Supports various paint types – Opaque Paints – Transparent Paints – Interference Paints 04/01/05 © 2005 University of Wisconsin
Rendering Limitations Kubelka-Munk doesn't account for: • Media of different refractive indices • Uniformly oriented pigment particles • Illumination other than diffuse • Fluorescent paints • Chemical or electrical interaction between different pigments Looks pretty good anyway… 04/01/05 © 2005 University of Wisconsin
Applications • “Interactive” painter • Semi-automatic “watercolorization” • NPR rendering (“watercolorization” in post) 04/01/05 © 2005 University of Wisconsin
Results 04/01/05 © 2005 University of Wisconsin
Results 04/01/05 © 2005 University of Wisconsin
Future Work • More effects – Spattering – Hairy brushes – Interaction with pen-and-ink • Fully automatic “watercolorization” – No manual masking – Find optimal palette • Generalization – Backruns and flow effects are really the same • Limit “shower door” effect in "watercolorized" animation. 04/01/05 © 2005 University of Wisconsin
Questions? 04/01/05 © 2005 University of Wisconsin
04/01/05 © 2005 University of Wisconsin
- Slides: 50