Image Segmentation A Hybrid Method Using Clustering

1
Image SegmentationA Hybrid Method Using
Clustering Region Growing

• Interim Presentation
• By Timothy Liao
• Supervisor Dr Sid Ray

2
Overview

• Image segmentation
• Research Context
• Image Segmentation Methods
• Clustering and Region Merging
• Implementation
• Conclusion

3
Image Segmentation

• What is it?
• First step in image analysis
• Partitioning of an image into non-overlapping
  regions.
• What is it used for?
• Detection of cancerous cells from medical images
• Detection of Roads from satellite images

4
Research Context

• Many Image Segmentation techniques.
• These techniques only work on certain images.
• Hybrid method, combining clustering and region
  merging.

5
Image Segmentation Methods

• Most image segmentation methods can be placed in
• one of three classes
• Characteristic feature thresholding or clustering
  (Feature Domain)
• Boundary detection (Spatial Domain)
• Region growing (Spatial Domain)

6
Clustering and Region Merging

• Clustering
• K-means Clustering
• ISODATA
• Fuzzy K-Means
• Region Merging
• Region Growing
• Split and Merge

7
K-Means Clustering

• K-means Clustering is Most common method used in
  unsupervised clustering.
• Prior knowledge of K is needed.
• Algorithm
• Select K different grey level values from pixels
  in an image.
• While K mean values ! previous k mean value
• do
• assign each pixel that has the closest grey
  level value to the k mean value.
• work out the new mean values for each k.
• end

8
Clustering of Grey Level Images
Original Image
K3
K2
We pick a K value to a apply to the image
depending on the analyser of the image.
9
Automatic Determination of K

• Ray and Turis automatic determination of K in
  colour image segmentation. (R.Tury, S.Ray 1998).
• Automatic determination of K for grey level
  images will be implemented in this research.

10
Region Merging

• Find Seed values within the image
• Merge pixels with similar grey level values
  together

Seeds
11
Region Merging

• Noise Removal
• Looking at spatial information to decide whether
  to merge a noise with the current region.

noise
12
Implementation

• Using C
• Using Monash Image Library
• Synthetic Images
• Natural Images

13
Conclusion

• Hybrid Image Segmentation technique should
  perform better than common techniques.

14
References

• R.H. Turi and S. Ray. K-means clustering for
  colour image segmentation with automatic
  detection of k. In Proceedings of Internation
  Conference on Sigmal and image Processing, pages
  345349, Las Vegas, Nevada, USA, 1998.
• M.R. Anderberg. Cluster Analysis for Application.
  New York Academic Press, 1973.
• J.T Tou and R.C. Gonzalez. Pattern Recognition
  Principles. Addison-Wesley., Massachusetts, USA,
  1974.
• E.W. Forgy. Cluster analysis of multivariate
  data eciency vs. interpretability of
  classifications. abstract, Biometrics,,
  21768769, 2000.
• J. MacQueen. Some methods for classification and
  analysis of multivariate observations. pages
  281279. Proceedings of Fifth Berkeley symposium
  on Mathematical Statistics and Probability, 1967.
• G. Coleman and H.C. Andrews. Image segmentation
  and clustering. pages 773785. Proc, IEEE, 1979.
• R E. Woods R C. Gonzalez. Digital Image
  Processing. Addison-Wesley, 1992.

15
Thank You

• Are there any questions or comments?