Contour Detection and Hierarchical Image Segmentation P. Arbelaez, - - PowerPoint PPT Presentation

contour detection and hierarchical image segmentation
SMART_READER_LITE
LIVE PREVIEW

Contour Detection and Hierarchical Image Segmentation P. Arbelaez, - - PowerPoint PPT Presentation

Contour Detection and Hierarchical Image Segmentation P. Arbelaez, M. Maire, C. Fowlkes, and J. Malik. IEEE TPAMI 2011. Islam Beltagy Original slides from: Hsin-Min Cheng http://archer.ee.nctu.edu.tw/powerpoint/GM_1024.pptx Outline


slide-1
SLIDE 1

Original slides from: Hsin-Min Cheng http://archer.ee.nctu.edu.tw/powerpoint/GM_1024.pptx Islam Beltagy

  • P. Arbelaez, M. Maire, C. Fowlkes, and J. Malik. IEEE TPAMI 2011.

Contour Detection and Hierarchical Image Segmentation

slide-2
SLIDE 2

Original slides from: Hsin-Min Cheng http://archer.ee.nctu.edu.tw/powerpoint/GM_1024.pptx

Outline

 Introduction  Contour Detection  Hierarchical Segmentation  Results and Evaluation

 Discussion

slide-3
SLIDE 3

Original slides from: Hsin-Min Cheng http://archer.ee.nctu.edu.tw/powerpoint/GM_1024.pptx

Contours (Ground Truth BSDS500)

Contour

slide-4
SLIDE 4

Original slides from: Hsin-Min Cheng http://archer.ee.nctu.edu.tw/powerpoint/GM_1024.pptx

Segmentation (Ground Truth-BSDS500)

slide-5
SLIDE 5

Original slides from: Hsin-Min Cheng http://archer.ee.nctu.edu.tw/powerpoint/GM_1024.pptx

Goal

  • Contour Detection
  • Segmentation from Contours

Original Image Segmentation Contour

slide-6
SLIDE 6

Original slides from: Hsin-Min Cheng http://archer.ee.nctu.edu.tw/powerpoint/GM_1024.pptx

Outline

 Introduction  Contour Detection  Hierarchical Segmentation  Results and Evaluation

 Discussion

slide-7
SLIDE 7

Original slides from: Hsin-Min Cheng http://archer.ee.nctu.edu.tw/powerpoint/GM_1024.pptx

Contours Detection

Original Image Contour = Pr(x, y, Ο)

slide-8
SLIDE 8

Original slides from: Hsin-Min Cheng http://archer.ee.nctu.edu.tw/powerpoint/GM_1024.pptx

  • Circular disk around (x, y) with orientation θ
  • Histograms of Intensities in both halves
  • ฀Chi Square distance between the two histograms
  • Three scales of r and eight values for θ

Contour Detection-Oriented gradient of histograms

slide-9
SLIDE 9

Original slides from: Hsin-Min Cheng http://archer.ee.nctu.edu.tw/powerpoint/GM_1024.pptx

  • Extract different channels
  • Brightness L, Color a, Color b
  • Components of CIE Lab colorspace
  • Texture

Contour Detection

Convolve K-mean clustering Replace pixel with cluster index

http://www.cs.berkeley.edu/~malik/papers/LM-3dtexton.pdf

slide-10
SLIDE 10

Original slides from: Hsin-Min Cheng http://archer.ee.nctu.edu.tw/powerpoint/GM_1024.pptx

  • Local Cues Combination

Contour Detection

Scale: 3 different values for the disc diameter α: learned from training data

max

slide-11
SLIDE 11

Original slides from: Hsin-Min Cheng http://archer.ee.nctu.edu.tw/powerpoint/GM_1024.pptx

  • Globalization

Contour Detection

Eigen Values Normalized Cuts Affinity Matrix Apply filters then sum

slide-12
SLIDE 12

Original slides from: Hsin-Min Cheng http://archer.ee.nctu.edu.tw/powerpoint/GM_1024.pptx

Outline

 Introduction  Contour Detection  Hierarchical Segmentation  Results and Evaluation

 Discussion

slide-13
SLIDE 13

Original slides from: Hsin-Min Cheng http://archer.ee.nctu.edu.tw/powerpoint/GM_1024.pptx

Uncertainty in Segmentation

Hierarchical Segmentation

slide-14
SLIDE 14

Original slides from: Hsin-Min Cheng http://archer.ee.nctu.edu.tw/powerpoint/GM_1024.pptx

Hierarchical Segmentation

 Oriented Watershed Transform  Ultrametric Contour Map

slide-15
SLIDE 15

Original slides from: Hsin-Min Cheng http://archer.ee.nctu.edu.tw/powerpoint/GM_1024.pptx

Oriented Watershed Transform

  • Watershed Transform

http://cmm.ensmp.fr/~beucher/wtshed.html

slide-16
SLIDE 16

Original slides from: Hsin-Min Cheng http://archer.ee.nctu.edu.tw/powerpoint/GM_1024.pptx

Oriented Watershed Transform

  • Watershed Transform

Artifacts

Weight each arc

slide-17
SLIDE 17

Original slides from: Hsin-Min Cheng http://archer.ee.nctu.edu.tw/powerpoint/GM_1024.pptx

Oriented Watershed Transform

WT OWT Fitting lines Weight arcs according to

  • rientation
slide-18
SLIDE 18

Original slides from: Hsin-Min Cheng http://archer.ee.nctu.edu.tw/powerpoint/GM_1024.pptx

Hierarchical Segmentation

  • Ultrametric Contour Map - Iterative Merging
slide-19
SLIDE 19

Original slides from: Hsin-Min Cheng http://archer.ee.nctu.edu.tw/powerpoint/GM_1024.pptx

Brief Summary

Original Image

Oriented Gradient

  • f histograms

Contour

Oriented Watershed Transform Ultrametric Contour Map

Hierarchical Segmentation

Threshold

slide-20
SLIDE 20

Original slides from: Hsin-Min Cheng http://archer.ee.nctu.edu.tw/powerpoint/GM_1024.pptx

Outline

 Introduction  Contour Detection  Hierarchical Segmentation  Results and Evaluation

 Discussion

slide-21
SLIDE 21

Original slides from: Hsin-Min Cheng http://archer.ee.nctu.edu.tw/powerpoint/GM_1024.pptx

Result (BSDS500 dataset)

Left to Right: Image, gPb-owt-ucm, threshold = optimal dataset scale, threshold = optimal picture scale

slide-22
SLIDE 22

Original slides from: Hsin-Min Cheng http://archer.ee.nctu.edu.tw/powerpoint/GM_1024.pptx

Result

Evaluation of segmentation algorithms Evaluation of contour detector

  • BSDS300 Dataset
slide-23
SLIDE 23

Original slides from: Hsin-Min Cheng http://archer.ee.nctu.edu.tw/powerpoint/GM_1024.pptx

Region benchmarks(2)

Covering Rand Index Variation of Information

  • ODS: optimal dataset scale
  • OIS: optimal image scale
slide-24
SLIDE 24

Original slides from: Hsin-Min Cheng http://archer.ee.nctu.edu.tw/powerpoint/GM_1024.pptx

Outline

 Introduction  Contour Detection  Hierarchical Segmentation  Results and Evaluation

 Discussion

slide-25
SLIDE 25

Original slides from: Hsin-Min Cheng http://archer.ee.nctu.edu.tw/powerpoint/GM_1024.pptx

Discussion

 What enforces closure in OWT-UCM for all

possible thresholds?

slide-26
SLIDE 26

Original slides from: Hsin-Min Cheng http://archer.ee.nctu.edu.tw/powerpoint/GM_1024.pptx

Discussion

 How to define ground truth ?

Can we use Hierarchal segmentation ?

slide-27
SLIDE 27

Original slides from: Hsin-Min Cheng http://archer.ee.nctu.edu.tw/powerpoint/GM_1024.pptx

Discussion

 OIS and the different possible segmentations

below, what do they suggest ?

slide-28
SLIDE 28

Original slides from: Hsin-Min Cheng http://archer.ee.nctu.edu.tw/powerpoint/GM_1024.pptx

Interactive Segmentation

1

slide-29
SLIDE 29

Original slides from: Hsin-Min Cheng http://archer.ee.nctu.edu.tw/powerpoint/GM_1024.pptx

Discussion

 Contours vs Segments.

slide-30
SLIDE 30

Original slides from: Hsin-Min Cheng http://archer.ee.nctu.edu.tw/powerpoint/GM_1024.pptx

Reference

  • P. Arbelaez, M. Maire, C. Fowlkes and J. Malik. Contour Detection

and Hierarchical Image Segmentation. IEEE TPAMI, Vol. 33, No. 5,

  • pp. 898-916, May 2011
  • P. Arbelaez, M. Maire, C. Fowlkes and J. Malik. From Contours to

Regions: An Empirical Evaluation. In CVPR 2009.

  • P. Arbelaez and L. Cohen. Constrained Image Segmentation from

Hierarchical Boundaries. In CVPR 2008.

slide-31
SLIDE 31

Original slides from: Hsin-Min Cheng http://archer.ee.nctu.edu.tw/powerpoint/GM_1024.pptx

Thanks