Expressing Rendering for Interactive Applications Recent results from two INRIA projects AERIS IPARLA A. Bousseau P. Barla J. Thollot X. Granier D. Vanderhaeghe R. Vergne (INRIA Grenoble) (INRIA Bordeaux)
Expressive Rendering for Interactive Applications Recent results from two INRIA projects AERIS IPARLA A. Bousseau P. Barla J. Thollot X. Granier D. Vanderhaeghe R. Vergne (INRIA Grenoble) (INRIA Bordeaux)
Goals ● New results in Expressive Rendering – Known as NPR (Non-Photorealistic Rendering) – Focus on user control and expressivity – Well adapted to video game industry. ● From two INRIA research teams – AERIS in Grenoble – IPARLA in Bordeaux
Presentation the two teams ● AERIS – INRIA Rhône-Alpes ● Universités de Grenoble - CNRS – Computer graphics – rendering – Previously known as ARTIS ● IPARLA – Computer graphics - 3D interaction / mobility – INRIA Bordeaux Sud-Ouest (01/01/2008) ● Université de Bordeaux - CNRS
Overview ● Classical Toon /Cell Shading – Color quantization – Static paper effects – Line drawing techniques ● Our solutions : more expressive shading – X-Toon – Dynamic canvas / Watercolorization – Dynamic drawing patterns / Shaded line drawings
Classic toon/cell shading
Toon Shading color quantization Simple shading Toon shading 0 1 0 1 N . L N . L
Toon Shading advantages/limitations ● Advantages – Extremely simple to implement – Very fast in both software and hardware – Fully compatible with any lighting system ● Limitations – Only one control : the 1D texture – Only based on the illumination ( 1D parameter ) – No paper or medium effect
Paper effect trivial solutions ● Paper/noise texture mapped on object surfaces – Looks like a wallpaper rather than a canvas property
Non-uniform frequency video
Paper effect trivial solutions ● Paper/noise texture mapped on object surfaces – Looks like a wallpaper rather than a canvas property ● Paper/noise texture applied to the whole image – Using texture sequences: no temporal coherence – Using a static texture: sliding, or shower door effect
Shower door effect video
Line drawing existing techniques ● Contours / Silhouettes – Object space: N.V 0 – Image space: Z discontinuities ● Ridges / Valleys – Object space: curvature extrema – Image space: Normal discontinuities
Line drawing advantages/limitations ● Advantages – Object and image space approaches – Can be combined to get more lines ● Limitations – Usually rendered directly, with no style – Patterns like hatching usually absent – Only geometric properties
Ours results
X-Toon [Barla et al. 2006] extended control ● New degree of freedom 1 – Using 2D textures depth – First dimension: ● lighting 0 1 lighting ● any other 1D attribute – Second dimension: 1 orientation ● depth/orientation ● any other 1D attribute lighting
X-Toon using depth 1 depth 0 1 lighting
X-Toon using depth : video video
Dynamic Canvas [Cunzi et al. 2003] paper effect ● Simulation of paper – grain / folding / ... – without shower door effect – with temporal coherence ● Principles – convert 3D to 2D ● camera movement to texture translation/rotation/zoom – infinite zoom ● adding/removing some frequencies
Dynamic Canvas video – 2D to 3D video
Dynamic Canvas video - results video
Watercolor effects ● Effects – non-uniform pigment density – edge darkening and wobbling – brush effects ● Causes – canvas and paper grain – water repartition – pigment flow and density – brush
Interactive Watercolor [Bousseau et al. 2006] Abstraction Quantization X-Toon
Interactive Watercolor [Bousseau et al. 2006] Wobbling Abstraction Quantization Offset based on gradient of paper texture
Interactive Watercolor [Bousseau et al. 2006] Contours Abstraction Wobbling Quantization Intensity change based on gradient of abstracted image
Interactive Watercolor [Bousseau et al. 2006] pigments/turbulence/paper Abstraction Wobbling Quantization Contours
Pigment and Turbulence temporal coherence ● Similar to dynamic canvas – convert 3D movement of object in 2D movements – multiple textures attached to particles ● particles on the object ● 2D translations transferred to the textures – infinite zoom on each texture
Pigment and Turbulence temporal coherence video
Interactive Watercolor [Bousseau et al. 2006] complete pipeline Abstraction Wobbling Quantization Contours medium
Watercolor results video
Watercolor current status ● Personalized style – expressivity of X-Toon – pigment / turbulence / paper choice ● Industrial transfer (with Studio Broceliande ) – new approach for a Maya plugin Graphanim – pending patent ● Extension to video stylization [Bousseau et al 2007] – abstraction using 2D + time segmentation – temporal coherence using optical flow http://www.studio-broceliande.com/
Dynamic 2D patterns [Breslav et al. 2007] ● Hatching / Stippling / ... – Uniform in 2D space – No surface parametrization ● Proposed approach – Subdivide into patches – Attach particles to the mesh – 3D to 2D local conversion – Blending between the patches
Dynamic 2D patterns [Breslav et al. 2007] Video http://code.google.com/p/jot-lib/ video
Shape depiction [in progress] information map ● Goals: – Preserving characteristic features – Style independancy ● Principle: using an information map – Lighting adjustment – Modification of light direction (exaggerated shading) – Extraction of lines
Shape depiction [in progress] using curvature map Information extraction object space (3D) Information map (curvature) X-Toon Rendering Objet 3D original (curvature direction) Processing, image space (2D)
Shape depiction [in progress] preliminary results X-Toon textures Ridges X-Toon black & white Exaggerated Lines Light adjustment Color adjustment
Other Results overview
Painterly rendering [ Vanderhaeghe et al. 2007] ● Brush strokes larger than 1 pixel ● Temporal Coherence – 3D to 2D movement conversion (particles) – preservation of a uniform density
Painterly rendering [ Vanderhaeghe et al. 2007] ● Brush strokes larger than 1 pixel ● Temporal Coherence – 3D to 2D movement conversion (particles) – preservation of a uniform density ● Style: user-defined strokes
Motif Synthesis [ Barla et al. 2006] ● Non-Uniform pattern – Similar to texture synthesis – Using vectorial information
Line simplification [ Barla et al. 2005] Clustering Line creation
Line simplification [ Barla et al. 2005]
Conclusion recent results ● Uniform 2D frequency – Patterns / paper / pigment / .... – Temporal coherency ● New styles – X-toon / dynamic pattern / watercolor / painterly – User controls – Expressivity – Interactive
Conclusion transfer / collaboration ● We need you – New research directions due to new problems ● We think that we can provide some solutions ● Different solutions – Direct partenership – National fundings (ANVAR, ANR, ....) – Consulting – PhD students or internships
Contacts and References ● Contacts – Pascal Barla pascal.barla@inria.fr – Xavier Granier xavier.granier@inria.fr – Joëlle Thollot joelle.thollot@inria.fr ● References Dynamic Canvas for Non-Photorealistic Walktroughs [Cunzi et al. 2003] – Geometric Clustering for Line Drawing Simplification [Barla et al. 2005] – Stroke Pattern Analysis and Synthesis [Barla et al. 2005] – Interactive Watercolor Rendering with Temporal Coherence and Abstraction [Bousseau et al. 2006] – X-Toon: An extended toon shader [Barla et al. 2006] – Dynamic 2D Patterns for Shading 3D Scenes [Breslav et al. 2007] – Dynamic point distribution for stroke-based rendering [Vanderhaeghe et al. 2007] – Video Watercolorization using Bidirectional Texture Advection [Vanderhaeghe et al. 2007] –
More Materials ● From NPR to Expressive Rendering ● X-Toon results using orientation ● Video – painterly rendering – line simplification – information map – watercolor for videos
From NPR to Expressive Rendering
Non-Photorealistic Rendering (NPR) ● Defined as the opposition to photorealism – Illumination laws are not the only rules ● Goal : « artistic » – Classical styles ● painting / drawing / engraving – New styles ● using the power of computers
Expressive Rendering ● Reveal different characteristics : – Shape / Materials – Illumination ● Goals : Expressiveness – User controls – Legibility – Depiction [Durand 2002]
X-Toon using orientation 1 orientation 0 1 lighting
Painterly Rendering video
Line simplification video
Information map video
Watercolorization of video video
Recommend
More recommend