Ballooning Snake Erika Tunyogi Feza Carlak Nicholas Popovic - - PowerPoint PPT Presentation

▶

Jun 04, 2023 441 likes •611 views

Introduction Active Contour Models - Theory Implementation Results and Outlook Ballooning Snake Erika Tunyogi Feza Carlak Nicholas Popovic Manuel Werlberger SSIP 2006 July 7, 2006 (SSIP 2006) Snake July 7, 2006 1 / 16 Introduction

SLIDE 1

Introduction Active Contour Models - Theory Implementation Results and Outlook

Ballooning Snake

Erika Tunyogi Feza Carlak Nicholas Popovic Manuel Werlberger

SSIP 2006

July 7, 2006

(SSIP 2006) Snake July 7, 2006 1 / 16

SLIDE 2

Introduction Active Contour Models - Theory Implementation Results and Outlook

Outline

1 Introduction 2 Active Contour Models - Theory 3 Implementation 4 Results and Outlook

(SSIP 2006) Snake July 7, 2006 2 / 16

SLIDE 3

Introduction Active Contour Models - Theory Implementation Results and Outlook

Overview

1 Introduction 2 Active Contour Models - Theory 3 Implementation 4 Results and Outlook

(SSIP 2006) Snake July 7, 2006 3 / 16

SLIDE 4

Introduction Active Contour Models - Theory Implementation Results and Outlook

Team Members

Erika Tunyogi

Theoretician, Researcher

Feza Carlak

Coordinator, Researcher

Nicolas Popovic

Graphic and Web Page Designer

Manuel Werlberger

Software Engineer

(SSIP 2006) Snake July 7, 2006 4 / 16

SLIDE 5

Introduction Active Contour Models - Theory Implementation Results and Outlook

Motivation

Initialisation Result

(SSIP 2006) Snake July 7, 2006 5 / 16

SLIDE 6

Introduction Active Contour Models - Theory Implementation Results and Outlook

Introduction

Snake is a form of Active Contour
Introduced by Kass, Witkin and Terzopoulous
Energy minimizing spline
Depending on Shape and Location of the Object
Used for Segmentation
Not sufficient to converge the contours to the image
Interaction process needed

(SSIP 2006) Snake July 7, 2006 6 / 16

SLIDE 7

Introduction Active Contour Models - Theory Implementation Results and Outlook

Overview

1 Introduction 2 Active Contour Models - Theory 3 Implementation 4 Results and Outlook

(SSIP 2006) Snake July 7, 2006 7 / 16

SLIDE 8

Introduction Active Contour Models - Theory Implementation Results and Outlook

Greedy Algorithm

The quantity is minimized by:

E =

(α(s)Econt + β(s)Ecurv + γ(s)Eimage)ds
Econt and Ecurv: first and second order continuity constraints
Eimage: measures image quantity (edge strength or intensity)

(SSIP 2006) Snake July 7, 2006 8 / 16

SLIDE 9

Introduction Active Contour Models - Theory Implementation Results and Outlook

Overview

1 Introduction 2 Active Contour Models - Theory 3 Implementation 4 Results and Outlook

(SSIP 2006) Snake July 7, 2006 9 / 16

SLIDE 10

Introduction Active Contour Models - Theory Implementation Results and Outlook

Implementation

Used Frameworks

Image Processing: Intel OpenCV library

http://sourceforge.net/projects/opencvlibrary/

GUI: Qt 4.1.4

http://www.trolltech.com/

Source Code Documentation: Doxygen

http://www.stack.nl/ dimitri/doxygen/

(SSIP 2006) Snake July 7, 2006 10 / 16

SLIDE 11

Introduction Active Contour Models - Theory Implementation Results and Outlook

Demo

(SSIP 2006) Snake July 7, 2006 11 / 16

SLIDE 12

Introduction Active Contour Models - Theory Implementation Results and Outlook

Algorithm

Iterative Algorithm
Neighborhood of each point is examined at each iteration
Energy calculation for the neighborhood
New location is determined by the minimized energy point
Only closed contours are considered

(SSIP 2006) Snake July 7, 2006 12 / 16

SLIDE 13

Introduction Active Contour Models - Theory Implementation Results and Outlook

Overview

1 Introduction 2 Active Contour Models - Theory 3 Implementation 4 Results and Outlook

(SSIP 2006) Snake July 7, 2006 13 / 16

SLIDE 14

Introduction Active Contour Models - Theory Implementation Results and Outlook

Results

(SSIP 2006) Snake July 7, 2006 14 / 16

SLIDE 15

Introduction Active Contour Models - Theory Implementation Results and Outlook

Outlook

More constraints could be added to obtain better results
Preprocessing steps could be implemented
3D Object Segmentation
Possibility to change parameters in the gui for different situations
Gradient Vector Flow (instead of Greedy Algorithm)
Switch from Snake to Balloons to cover concave structures

Additional Information:

SSIP2006 - Project Homepage
http://snake.sourceforge.net/

(SSIP 2006) Snake July 7, 2006 15 / 16

SLIDE 16

Thank you very much for your attention!

(SSIP 2006) Snake July 7, 2006 16 / 16