SLIDE 1
Outline
- Overview
- Constrained Parametric Min-Cuts (CPMC)
– Experiments
- Example Images
- Distorted Images
- Ranking Object Hypotheses
– Experiments
- Depth/Disparity Cues
- Discussion
2
Automatic Object Segmentation Joao Carreira and Cristian - - PowerPoint PPT Presentation
Automatic Object Segmentation Joao Carreira and Cristian - - PowerPoint PPT Presentation
Constrained Parametric Min-Cuts for Automatic Object Segmentation Joao Carreira and Cristian Sminchisescu Presenter: Che-Chun Su 2012/09/28 Outline Overview Constrained Parametric Min-Cuts (CPMC) Experiments Example Images
SLIDE 2
SLIDE 3
Overview
3
Figure credit: Joao Carreira et al.
SLIDE 4
Constrained Parametric Min-Cuts (CPMC)
- Graph-based segmentation algorithm
– Similarity between neighboring pixels is encoded as edges.
4
SLIDE 5
Constrained Parametric Min-Cuts (CPMC)
- Multi-Cue Contour Detector
– Estimate the posterior probability of a boundary.
5
Figure credit: Michael Maire et al.
SLIDE 6
Experiments
- Segmentation Covering
6
SLIDE 7
Experiments
- Example Images
7
SLIDE 8
Experiments
- Example Images
8
SLIDE 9
Experiments – Distorted Images
- Will different distortions in images affect the
segmentation performance?
- Will the distortion degrade the quality of the
estimated posterior probability of boundary?
- LIVE Image Quality Database
– Gaussian blur – JPEG compression – White noise
9
SLIDE 10
Test Images
10
Reference Blur JPEG White Noise
SLIDE 11
Probability of Boundary Map
11
Reference Blur JPEG White Noise
SLIDE 12
Experiments
12
- Reference
SLIDE 13
- Blur
Experiments
13
SLIDE 14
- JPEG
Experiments
14
SLIDE 15
- White Noise
Experiments
15
SLIDE 16
Ranking Object Hypotheses
16
Figure credit: Joao Carreira et al.
SLIDE 17
Experiments
- Can depth cues help rank the object hypotheses?
– Depth are continuous; however, objects can be seen as residing in different depth planes.
- Middlebury Stereo Datasets
– Ground-truth disparity maps
- LIVE Color+3D Database
– Ground-truth range maps
17
SLIDE 18
Experiments
- Append the feature with depth/disparity cues and retrain the
ranking model with multiple linear regression.
18
SLIDE 19
Experiments
- Middlebury Stereo Datasets
– Indoor scenes with ground-truth disparity maps – Different types of objects – Ranking model is trained on LIVE Color+3D database.
19
SLIDE 20
Experiments
20
SLIDE 21
Original Features
21
SLIDE 22
New Features and Regressor
22
SLIDE 23
Original Features
23
SLIDE 24
New Features and Regressor
24
SLIDE 25
Original Features
25
0.196388 0.452087 0.505323 0.615173
SLIDE 26
New Features and Regressor
26
0.196388 0.424314 0.490003 0.450192
SLIDE 27
Experiments
- LIVE Color+3D Database
– Natural scenes with ground-truth range maps – Quantize actual range values to generate depth planes. – Ranking model is trained on Middlebury stereo datasets.
27
SLIDE 28
Experiments
28
SLIDE 29
Experiments
29
SLIDE 30
Original Features
30
SLIDE 31
New Features and Regressor
31
SLIDE 32
Original Features
32
0.407832 0.337091 0.133830 0.187111
SLIDE 33
New Features and Regressor
33
0.407832 0.333177 0.133830 0.179389
SLIDE 34
Discussion
- Different types of distortions in images can affect the
segmentation results.
– Probability of boundary map is distorted. – CPMC generates incorrect figure-ground (object) hypotheses.
- Ranking model can be governed by different types of
segment features and properties.
– Depth cues could possibly help recognize objects, and vice versa.
34