Stroke-Based Rendering - Advanced Computer Graphics - Lecture Slides, Slides of Computer Graphics

Download Stroke-Based Rendering - Advanced Computer Graphics - Lecture Slides and more Slides Computer Graphics in PDF only on Docsity!

Non-Photorealistic Rendering

Tobias Isenberg

Stroke-Based Rendering

Overview

• introduction and overview

• optimization strategies

• greedy algorithms

Introduction

• stroke as the main primitive

• (semi-)automatic placement of discrete elements
• use of objective function to measure placement quality
• typical goal: represent an image or a 3D shape
• result: abstracted version of the source image/shape

Introduction

• control of stroke placement

• automatic
• semi-automatic
• interactive

• human control at all levels

• decision on the stroke set
• decision on the source image
• parameters of objective function
• semi-automatic stroke placement
• interactive stroke placement

© Gavin Ross, (http://www.gavinross.com/)used with permission

Definitions

• necessary elements for stroke-based rendering:

• canvas: background color/texture/object
• ordered list of strokes with parameterization, to be rendered (alpha-blended) onto the background
• SBR energy function: measurement of image quality

E: I  R; I = set of possible images, R = real numbers

Energy Function E(I)

• measures how closely input image is matched

• also encodes trade-offs

• e.g., abstraction by enforcing larger strokes
• otherwise very many (m×n) tiny paint strokes (pixels)

• example for an SBR energy function

• wabs is a factor to parameterize the influence of Eabs
• S is the source image

E I I

E I x y S x y

E I E I w E I

abs

xy I

match

match abs abs

( ) number of strokes in

( , )

2



Stroke-Based Rendering Strategies

• optimization strategies

• Voronoi algorithms: Lloyd’s method and variations
• trial-and-error algorithms

• greedy algorithms

• rendering in one go

Stroke-Based Rendering

Optimization Strategies

Voronoi Algorithms

• general class of applications

• images that contain many non-overlapping strokes
• only stroke density is constrained

• examples

• stippling: small black dots
• mosaics: small colored pieces

• general idea

• use efficient algorithms from computational geometry
• place elements evenly, no overlaps
• use GPU to speed up the process

• note: energy function defined as density, not tone

Lloyd’s Method: Brief Recap

• iterative optimization technique

• minimize energy function describing the distance

between points and their Voronoi region centroids

• estimation of Voronoi regions can be done on GPU

• previously seen two techniques

• interactive techniques using brushes
• automatic technique with weighted Voronoi regions

Lloyd’s Method to Create Mosaics

• modifications to Lloyd’s method:

• use of Manhattan distance to obtain rectangular tiles
• use an underlying vector field to guide the computation of this metric: derived from the image’s edge information
• computing the Voronoi diagrams through rendering: pyramids at point positions with ID color, orthographic
• results in discrete Voronoi diagram

Lloyd’s Method to Create Mosaics

• extension of the basic technique

• avoiding an edge by affecting the centroid computation: rendering non-ID color over Voronoi diagram affects the centroid computation so that eventually tiles avoid it
• result: rendering of image edges included, tiles avoid features

Lloyd’s Method to Create Mosaics

Lloyd’s Method to Create Mosaics