Segmentation of the glottal space from laryngeal images using the watershed transform

1
Segmentation of the glottal space from laryngeal
images using the watershed transform

• Authors Victor Osma-Ruiz , Juan I.
  Godino-Llorente, Nicolas Saenz-Lechon, Ruben
  Fraile
• Source Computerized Medical Imaging and Graphics
• speaker Kuo Hua Wang

2
Outline

• Introduction
• Proposed scheme
• Experimental results
• Conclusions

3
Introduction
Glottal space
Original Gray Scale Image
Watershed Transform
Thresholded Gradient
Surrounding Regions Merging
Image with the segmented Glottis
Decision Making
JND Merging
4
Proposed scheme
luminance component (Y)
Thresholded Gradient
those pixels of the gradient image with a grey
level below 2 are assigned to 0.
5
Proposed scheme
Watershed Transform
6
Proposed scheme
JND Merging
mR1 mR2 mean values of 2 regions
mR1 100 mR2 150 R1 100 R2 1000
7
Proposed scheme
Surrounding Regions Merging
8
Proposed scheme
Decision Making
9
Experimental results

• The methodology has been tested with 111 images,
  taken from 15 videos.
• The mistakes appear in 5 of the images analyzed.

10
Conclusions

• A new method for the automatic segmentation of
  the glottal space has been presented
• The solution presented is very promising.