Title: Segmentation and Boundary Detection Using Multiscale Measurements
1Segmentation and Boundary Detection Using
Multiscale Measurements
Ronen BasriAchi BrandtMeirav GalunEitan Sharon
2Image Segmentation
3Local Uncertainty
4Global Certainty
5Local Uncertainty
6Global Certainty
7Coarse Measurements for Texture
8A Chicken and Egg Problem
Problem Coarse measurements mix neighboring
statistics
Solution support of measurements is determined
as the segmentation process proceeds
9Segmentation by Weighted Aggregation
- Normalized-cuts measure in graphs
- Complete hierarchy in linear time
- Use multiscale measures of
- intensity, texture, shape, and boundary
integrity
10Segmentation by Weighted Aggregation
- Normalized-cuts measure in graphs
- Complete hierarchy in linear time
- Use multiscale measures of
- intensity, texture, shape, and boundary
integrity
11Segmentation by Weighted Aggregation
- Normalized-cuts measure in graphs
- Complete hierarchy in linear time
- Use multiscale measures of
- intensity, texture, shape and boundary
integrity
12The Pixel Graph
Couplings Reflect intensity similarity
Low contrast strong coupling
High contrast weak coupling
13Hierarchical Graph
14Hierarchyin SWA
15Normalized-Cut Measure
16Normalized-Cut Measure
Minimize
17Normalized-Cut Measure
Low-energy cut
Minimize
18Segment Detection
19Coarse-Scale Measurements
- Average intensities of aggregates
- Multiscale intensity-variances of aggregates
- Multiscale shape-moments of aggregates
- Boundary alignment between aggregates
20Adaptive vs. Rigid Measurements
Original
Averaging
Geometric
Our algorithm - SWA
21Adaptive vs. Rigid Measurements
Original
Interpolation
Geometric
Our algorithm - SWA
22Use Averages to Modify the Graph
23Use Averages to Modify the Graph
24Texture Examples
25Isotropic and Oriented Filters
A brief tutorial
Textons by K-Means Malik et al IJCV2001
26Oriented Texture in SWA
with Meirav Galun
Shape Moments
- center of mass
- width
- length
- orientation
Oriented Texture of aggregate orientation,
width and length in all scales
27Boundary Integrity in SWA
28Hierarchyin SWA
29SWA
Detects the salient segments
Hierarchical structure
Linear in of points (a few dozen operations per
point)
30Experiments
- Our SWA algorithm (CVPR00 CVPR01)
- run-time 5-10 seconds.
- Normalized cuts (Shi and Malik, PAMI00 Malik
et al., IJCV01) - run-time about 10-15 minutes.
- Software courtesy of Doron Tal, UC Berkeley.
31Isotropic Texture - Horse I
Our Algorithm (SWA)
Normalized Cuts
32Isotropic Texture - Horse II
Our Algorithm (SWA)
Normalized Cuts
33Isotropic Texture - Tiger
Our Algorithm (SWA)
Normalized Cuts
34Isotropic Texture - Butterfly
Our Algorithm (SWA)
Normalized Cuts
35Isotropic Texture - Leopard
Our Algorithm (SWA)
36Isotropic Texture - Dalmatian Dog
Our Algorithm (SWA)
37Isotropic Texture - Squirrel
Our Algorithm (SWA)
Normalized Cuts
38Full Texture - Squirrel
Our Algorithm (SWA)
Normalized Cuts
with Meirav Galun
39Full Texture - Composition
Our Algorithm (SWA)
with Meirav Galun
40Full Texture Lion Cub
Our Algorithm (SWA)
with Meirav Galun
41Full Texture Polar Bear
Our Algorithm (SWA)
with Meirav Galun
42Full Texture Penguin
Our Algorithm (SWA)
with Meirav Galun
43Full Texture Leopard
Our Algorithm (SWA)
with Meirav Galun
44Full Texture Leopard
Our Algorithm (SWA)
with Meirav Galun
45Full Texture Owl
Our Algorithm (SWA)
with Meirav Galun
46Full Texture Bird
Our Algorithm (SWA)
with Meirav Galun
47Separation of Parts
Poissonian u ?u 1 u 0 outside the segment