Active Strokes: Coherent Line Stylization for Animated 3D Models

ACTIVE STROKES:COHERENT LINE STYLIZATION FOR ANIMATED 3D MODELS

Pierre Bénard1,2 Jingwan Lu3 Forrester Cole4 Adam Finkelstein3 Joëlle Thollot1,2

1Grenoble University, LJK 2INRIA 3Princeton University 4MIT CSAIL

2

Line Stylization

3

Line Stylization in CG

4

Line Stylization in CG• Image space buffers

processing

Comprehensible rendering of 3-D shapes,Saito and Takahashi, SIGGRAPH 1990

• Object space line extraction

Real-Time Nonphotorealistic Rendering,Markosian et al., SIGGRAPH 1997

5

Line Stylization: Image Space

Implicit Brushes for stylized line-based rendering, Vergne et al., CGF 2011

Simple and fastNatural coherence and LoD

Restricted stylization effects

6

Line Stylization: Object SpaceWide range of styles Computationally expensive

Complex LoD

No natural coherence

A Procedural Approach to Style for NPR Line Drawing from 3D models,

Grabli et al., TOG 2010

7

Brush Path Parameterization

0

l

Brush path

Line texture

Feature lines

8

Creases Ridges and valleys

Silhouettes Suggestive contours Apparent ridges

Fixed Line View-Dependent Lines

Parameterization = arc-length Parameterization = ?

9

Coherent Parameterization Goals

• Flatness Remain linear in image space

• Motion coherence Evolve according to the motion of the object

• Temporal continuity Adapt to the topological events

10

Previous Work

Coherent Stylized Silhouettes, Kalnins et al., SIGGRAPH 2003

Optimization between 2D and 3D

Multiple brush strokes per line

Dependent on the input connectivity

11

Previous Work

12

Previous Work

Kalnins et al., Coherent Stylized Silhouettes, SIGGRAPH 2003

13

Previous Work

Self-Similar Texture for Coherent Line Stylization, Bénard et al. NPAR 2010

2D Infinite zoom:Self-Similar Line Artmap (SLAM)

Dependent on the input connectivity

One brush stroke per line

14

Previous WorkEase propagation

of parameterizationwith CSS

Input connectivity

Snaxels on a Plane,Kevin Karsch and John C. Hart,

NPAR 2011

Spatio-Temporal Analysis for Parameterizing Animated Lines,

Buchholz et al., NPAR 2011

Optimization over the entire animation

Offline computation

15

Our Approach: Active Strokes• Image space active contours [Kass et al. 1988]

Shape Topology Density

• Brush paths Coherent parameterization Shape abstraction

• Interactive frame rates

16

SnakesTracking

Pipeline

Advection

3D scene

Image space lines extraction

Feature samples

Relaxation

Vectorization

Coverage Connectivity

Line drawing

Brush PathsGeometry

Parameterization

Stylization

17

Input• Feature samples extracted in image space

2D position Local tangent 2D velocity

18

SnakesTracking

Pipeline

Advection

3D scene

Image space lines extraction

Feature samples

Relaxation

Vectorization

Coverage Connectivity

Line drawing

Brush PathsGeometry

Parameterization

Stylization

19

Snakes: Properties

Coherence continuous evolution across frames

Accuracy faithfully representation of shape

Coverage level of detail Simplicity simple topology Length stylization freedom

Trac

king

Vect

oriz

ation

20

Active Contours in NPR• Color regions tracking

in videos Off-line computation User corrections

SnakeToonz, AgrawalaNPAR 2002

Keyframe-Based Tracking for Rotoscoping and Animation,

Agrawala et al. SIGGRAPH 2004

21

Snakes: Tracking• Advection + Relaxation

Frame f Frame f+1

22

Snakes: Advection• Reprojection

similar to Bousseau et al. 2007 and Lu et al. 2010

Image fImage f+1

23

Snakes: Relaxation [Kass et al. 1988]

• Minimize the energy:

• Semi-explicit Euler scheme

ExternalAttraction by the features samples

Internal– Continuity– Smoothness

24

Snakes: Relaxation• Grow / shrink

re-sampling (similar to Delingette et al. 2000)

• Contour shrinking mass-spring forces between the vertices almost constant length if no tangential external force is applied

25

Snakes: Tracking• Advection + Relaxation

26

Snakes: Cleaning• Local vectorization operators

• Applied sequentially in a greedy fashion

27

Snakes: Tracking + Cleaning

28

Snakes: Level of Detail

29

SnakesTracking

Pipeline

Advection

Feature samples

Relaxation

Vectorization

Coverage Connectivity

Brush PathsGeometry

Parameterization

Stylization

3D scene

Image space lines extraction

Line drawing

30

Brush Paths: Parameterization• Linear image space parameterization:

with = slope= phase= arc-length

• Evolve according to the motion and topology of the snakes parameterization at each vertex

31

Brush Paths: Parameterization• Propagation at each vertex

Parameterization stored between two frames

32

Brush Paths: Parameterization• Propagation at each vertex• Linearization (in the least-square sense)

para

met

eriza

tion

arc-length

33

Brush Paths: Parameterization• Propagation at each vertex• Vectorization events

Split Extend Trim Merge: mechanism to avoid parameterization

discontinuities

propagated directly

• Linearization

34

Brush Paths: Merging

• Only if the slope and phase match• Leveling mechanism

pushes the 2 parameterizations to their mid-value:

with

para

met

eriza

tion

arc-length

35

Stylization: SLAM texture

36

Stylization: SLAM offsets

37

Stylization: Shape abstraction• Segments fitting

38

Stylization: Shape abstraction• Arcs fitting

39

Stylization: Shape abstraction

40

Stylization: Implicit Brushes

41

Stylization: Steerable filters

42

Stylization: Deformable models

43

Stylization: Deformable models

44

Limitations & Future Work

Robustness toward noise / outliers Temporal smoothing

Input samples Active strokes

45

Conclusions• New propagation primitive

for image space lines fully automatic running at interactive frame rates

• Temporally coherent basis for complex effects Arc-length variations (texture, offsets, tapering)

Shape abstraction

QUESTIONS?Thank you for your attention

46