Level Set Methods An Overview - PowerPoint PPT Presentation

1 / 28
About This Presentation
Title:

Level Set Methods An Overview

Description:

L = Local Properties (curvature, normal direction) G ... FA is called the advection speed function ... Advection. The domain of dependence of point A is point B ... – PowerPoint PPT presentation

Number of Views:96
Avg rating:3.0/5.0
Slides: 29
Provided by: amerab
Category:

less

Transcript and Presenter's Notes

Title: Level Set Methods An Overview


1
Level Set Methods(An Overview)
  • Amer Abufadel
  • November 10, 2000

2
Propagating Interfaces
  • Ocean Waves
  • Burning Flames
  • Material Boundaries
  • Shapes against background
  • Handwriting recognition
  • Image Contours and segmentation

3
Overview
  • Interface Propagation
  • Boundary Value Problem
  • Initial Value Problem
  • Applications
  • Results

4
Interface Propagation
Inside
Outside
Outside
5
Interface Propagation
  • Speed Function F depends on
  • L Local Properties (curvature, normal
    direction)
  • G Global Properties (pressure, heat source)
  • I Independent Properties (fluid velocity)

6
One extra dimension
7
Boundary Value Formulation
T
T 2
Y
T 1
T 0
T 1
T 0
T 2
X
8
Boundary Value Formulation
  • Assume F gt 0
  • Front Moving Outward
  • Crossing Time at
  • (x,y) for expanding front F gt 0

9
Initial Value Formulation
F(x,y,t2)
F0
F(x,y,t1)
F0
F(x,y,t0)
F0
10
Initial Value Formulation
  • F positive or negative
  • Boundary can cross point more than once
  • T(x,y) is not single valued
  • Make the initial position as a zero level of a
    higher-dimensional function f
  • Track evolution of f and determine the zero level
    set

11
Advantages
  • Unchanged in higher dimensions
  • Topological changes are handled naturally
  • Accurately approximated by computational schemes
  • Intrinsic geometric properties are easily
    calculated
  • Efficient

12
Advantages
  • Immune to topological changes

13
Level Set Approach
  • Boundary considered as the zero value of
    hypersurface
  • As hypersurface moves under a speed function, the
    boundary moves as well
  • Image based speed function will stop evolution
    near boundaries

14
Speed Function
  • FA is called the advection speed function
  • FG is called the geometric speed function and is
    dependent on curvature of the surface

15
Advection
t
A
x
B
  • The domain of dependence of point A is point B
  • The domain of influence of point B is the set of
    points on line A,B

16
Sample Speed Function
17
Speed Function Extension
  • Speed function derived from the image is only
    valid at the boundary,i.e. where the function has
    zero value.
  • What do we do about other points on the surface?
    Extend the speed function

18
Speed Function Extension
19
Speed Function Extension
  • For every point P on the grid
  • Find closest point on boundary Q
  • Assign the value of the image based speed of the
    point Q to point P

20
The Level Set Algorithm
  • Start with an initial condition (position of the
    front at t0)
  • Repeat until done
  • Calculate the extension velocity
  • Calculate the surface at time n1 from the
    current time n
  • Calculate the level set at zero

21
Level Set Method
  • Update equation

22
Level Set Method
The update equation can be simplified for easy
programming
where
Here
23
Practical Issues
  • Algorithm is very computationally intensive
  • Can be done in a more efficient manner
  • Narrow Band Level Set Method
  • Fast Marching Methods

24
Narrow Band Level Set Method
  • The efficiency comes from updating the speed
    function.
  • We do not need to update the function over the
    whole image or volume.
  • Update over a narrow band (2D) or tube (3D)

25
Narrow Band Level Set Method
  • Update over a narrow band (2D) or tube (3D)

Update Band
26
Results
  • http//math.lbl.gov/malladi/

27
My Research Interest
  • Apply level sets to PET and SPECT reconstructions
  • Isolate the blood pool
  • Accurate ejection fraction calculation

28
References
  • Sethian J.A. Level Set Methods and Fast Marching
    Methods, Cambridge University Press, Cambridge
    UK, 1999
  • Malladi R., Sethian J.A., and Vemuri B.C., Shape
    Modeling with Front Propagation A Level Set
    Approach IEEE Trans. On Pattern Analysis and
    Machine Intelligence, 17, , pp.158-175, 1995
  • http//math.lbl.gov/malladi/
Write a Comment
User Comments (0)
About PowerShow.com