Stylization and Abstraction of Photographs

1
Stylization and Abstraction of Photographs

• Doug DeCarlo
• Anthony Santella

2
Outline

• Introduction
• Related Work
• Background
• Overview
• Hierarchical Image Representation
• Rendering with a Perceptual Model
• Results
• Conclusion

3
Introduction

• Good information design
• Reduce the effort required to understand an image
• Abstraction
• A tool for effective visual communication

4
Introduction

• Clarify the meaningful visual structure

5
Introduction

• Represent image as hierarchical structure of
  parts and boundaries
• Apply perceptual model with data from an eye
  tracker

6
Introduction

• Non-Photorealistic Rendering (Line Drawing)

7
Related Work

• Line-Drawing Style Curtis 1999
• Selective Omission
• Illumination model Gooch et al. 1998
• Omit texture information Kowalski et al. 1999
  Deussen and Strothotte 2000

8
Related Work

• Abstracted Painterly Renderings Using
  Eye-Tracking Data Santella and Decarlo 2002

9
Related Work

• Abstracted Painterly Renderings Using
  Eye-Tracking Data Santella and Decarlo 2002

10
Background

• Edge detection Meer and Georgescu 2001
• Image Segmentation Comaniciu and Meer 2002
• Combined into a single system Christoudias et
  al. 2002
• Scale-Space Theory Koenderink 1984 Lindeberg
  1994

11
Background

• Visual Perception Model
• Fixations/Saccades
• Eccentricity
• Contrast sensitivity

12
Overview

• Obtain Eye Movement Data
• Represent Image in Hierarchical Form
• Render Image with Perceptual Model

13
Hierarchical Image Representation

• Create Image Pyramid Burt and Adelson 1983
• Compute Segmentation Comaniciu and Meer 2002
• Detect Edges Meer and Georgescu 2001
• Edge Tracking Truco and Verri 1998
• Build the Hierarchy

14
Hierarchical Image Representation

• Create Image Pyramid Burt and Adelson 1983
• Compute Segmentation Comaniciu and Meer 2002

15
Hierarchical Image Representation

• Detect Edges Meer and Georgescu 2001
• Edge Tracking
• Truco and Verri 1998

16
Hierarchical Image Representation

• Building the Hierarchy (Segmentation Tree)

17
Hierarchical Image Representation

• Building the Hierarchy (Segmentation Tree)

level L
18
Hierarchical Image Representation

• Building the Hierarchy (segmentation tree)

19
Hierarchical Image Representation

• Building the Hierarchy (segmentation tree)

20
Hierarchical Image Representation

• Building the Hierarchy (segmentation tree)

21
Hierarchical Image Representation

• Building the Hierarchy (segmentation tree)

22
Rendering with a Perceptual Model

• Prune Segmentation Tree
• Smooth Region Boundaries
• Finkelstein and Salesin 1994
• Draw Regions
• Overlay Lines

23
Rendering with a Perceptual Model

• Draw regions on a particular frontier
• Prune the
• segmentation tree

24
Rendering with a Perceptual Model

• Pruning the Segmentation Tree
• Find target region (ni) for fixation (fi)
• Compute region eccentricity ei
• with respect to fi
• Apply SPLIT(n) to regions in depth first search
  order

25
Rendering with a Perceptual Model

• Pruning the Segmentation Tree
• Region is perceptible when

26
Rendering with a Perceptual Model

• Smoothing Boundaries Finkelstein and Salesin
  1994
• Filter with Gaussian kernel

27
Rendering with a Perceptual Model

• Draw Regions (Coarse-to-Fine Order)
• Draw Lines
• Line is visible if
• Lines must lie along the frontier boundary
• Thickness depends on length
• Linearly taper at each end

28
Results
29
Results
30
Results
31
Results
32
Conclusion

• Contribution
• Formulated the process of abstraction in terms of
  a model of visual form
• Future Work
• Apply model of shading (remove patchiness)
• Texture
• Additional aspects of visual form
• Animation