Scribd
Upload
94 views7 pages

Graph-Based Segmentation - PPT Download

Uploaded by

Shailesh Reddy
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
94 views7 pages

Graph-Based Segmentation - PPT Download

Uploaded by

Shailesh Reddy
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 7

Log in 

Search... Search

2 / 29

Similar presentations
Graph-based Segmentation 

Published by Pauline Reynolds Modified over 5 years ago

 Embed

 Download presentation
9
Presentation on theme: "Graph-based Segmentation"—
Presentation transcript:

1 Graph-based Segmentation
2 Main Ideas Convert image into a graph
Vertices for the pixels
Edges between the pixels
Additional vertices and edges to encode other constraints
Manipulate the graph to segment the image

3 Papers
Interactive graph cuts for optimal boundary & region segmentation of
objects in N-D images
Boykov and Jolly
Minimize an energy function
Efficient Graph-based Segmentation
Felzenszwalb and Huttenlocher
Cluster the vertices based on edge weight

4 Boykov and Jolly Binary image segmentation Minimise an


energy function
Classify pixels as object or background
Their contribution is adding interactivity
Minimise an energy function
E(A) = B(A) + λR(A)
A: Segmentation (assign each pixels to object or background)
B(A): The cost of all the edges between object pixels and background
pixels
R(A): The cost of deciding a pixel to be object or background

5 Creating the Graph Each pixel has a corresponding vertex


Additionally, a source (“object”) and a sink (“background”)
Each pixel vertex has an edge to its neighbours (e.g. 8 adjacent
neighbours in 2D), an edge to the source, an edge to the sink

6 Edge Weights between pixels


Weight of edges between pixel vertices are determined by the B()
function
Low score when boundary is likely to pass between the vertices
high score when vertices are probably part of the same element
E.g. the difference in pixel intensities, the gradient

7 Edges to Source/Sink
If pixel is known to be an object, use a high weight (K) to the source, zero
weight to the sink
K is chosen so that it will never be cut
Conversely, if pixel is background, use weight K to the sink, zero weight
to the source
Otherwise, weigh edges to source and sink appropriately using the R()
function
Note that the edge to the source is the “likelihood” for the pixel being the
background – we break this edge when the pixel is assigned to the
background

8 Applications Handles arbitrary number of dimensions


Finds global minimum energy
Needs “good” user input to work effectively
Need intelligent functions
Need to select λ

9 Felzenszwalb and Huttenlocher


Download the program from the webpage:
Minimal documentation
Short README file
Paper

10 The program Comes as a tar.gz archive and .zip archive


Process
Extract archive
‘make’ (Makefile supplied)
Program consists of
A .cpp “wrapper” file (only calls the functions)
Actual algorithm functions are in .h files
Basic portable C++ code

11 Program Testing Built on Mac OS X, Linux, Windows cygwin


Gcc toolchain, but any C++ compiler should work
Supplied basic Makefile
Results were basically the same between platforms
Colors are chosen randomly
Results obtained are not the same as posted on the website
Image files on the website may be modified
(scaled/compressed/downsampled)

12 Algorithm Create a graph Edges are between “neighbouring”


vertices
Each vertex corresponds to a vertex
Edges are between “neighbouring” vertices
Choose a small neighbourhood to reduce computation time (otherwise
we have a complete graph)
Weight on the edge is the 5D distance between the points (for a 2D
image)
5D vector = x position, y position and 3 color components

13 Parameters
σ: Use this value and do Gaussian smoothing (preprocessing the image
to reduce noise)
k: threshold value for doing the clustering
min: “hack” parameter
the smallest cluster size must contain at least this many vertices –
clusters that are too small will be merged with other clusters until
sufficiently large
14 Clustering Put each vertex in a component Sort edges by
weight
Take each edge in turn
If the edge is between vertices in two different components A and B, we
can merge if the edge weight is lower enough than the threshold
Threshold is the minimum of the following value, computed on A and B
(Lowest weight edge in minimum spanning tree of the component) + (k /
size of component)

15 Notes
Low edge weights between vertices that are likely to be in the same
cluster
As a cluster gets larger, it becomes harder to add vertices to it
Heuristic – not really minimising a particular energy function
More similar to “region growing”
User has select “good” parameters to get good results

16 Effect of σ Increased smoothing results in removal of noise


Can cause “bleeding” – the algorithm has difficulty separating
background from the object if the boundaries are too smooth

17 Reference Images
18 Increasing σ Clouds are recognised as one object
Palm tree gets confused with ocean

19 Grain
Increasing σ introduces more blurring (reduces the edge weight between
pixels)

20 Vertebrae MRI
Gets rid of noise (bottom left, right hand side), but purple vertebrate
piece bleeds out

21 Increasing Threshold Clusters more aggresively


Palm tree is confused with ocean and clouds

22 Grain Non-grain pixels are almost all clustered together


Measure of how “similar” all the pixels of an object should be

23 MRI
Vertebrae and region next to vertebrae are very similar shade so are
easily confused

24 Increasing ‘min’ value


Segmentation is the same, but small components are merged with
neighboring ones

25 Grain
Easy to control change, gets rid of small artifacts
26 MRI
Not much effect if regions are already large

27 Parameter Tweaking
Need to manually tune parameters to get a good image
Image of MRI after selecting
σ = 0.6
k = 200
min = 60

28 Performance Need to tune parameters by hand Very fast


Program usually takes a couple seconds on the test images provided
Only takes ppm images in RAW data format
Theoretically, algorithm generalises to arbitrary number of dimensions
and arbitrary number of features per pixel

29 Command-line tools
Working with images involves opening up the results in an image viewer
– can get messy

Download ppt "Graph-based Segmentation"

 Similar presentations
© 2022 SlidePlayer.com Inc. Feedback Do Not Sell About project
All rights reserved.
Privacy Policy My Personal SlidePlayer
Feedback Information Terms of Service

Search... Search

You might also like