Jeremy S. De Bonet : Texture Driven Segementation




Resume
RESEARCH
     
Publications
Image Compression
Texture Synthesis
Image Database Retrieval
SEGMENTATION
Registration
Discrimination
Projects
Web Hacks


SAR Target Segmentation

For synthetic aperature radar targets, chips were taken from the MSTAR CDs and one 16x16 training example was used for (each of) the grass, vehicle and shadow textures. (Of course, the same training examples were used for all the pictures below.)

In each picture, the frames are (in reading order):

  • Original image (histogram equalized)
  • Segmented grass region
  • Segmented vehicle region
  • Segmented shadow region
  • Initial maximum likelihood segmentation mask
  • Smoothness prior applied to segmentation mask

No optiminzation was done to choose example textures. Doing so could make it better. Also, multiple examples of each texture class could be used.


Knee MRI Segmentation

The MRI image of a knee was segmented using a 10-way texture classification. The bright colors indicate anatomical regions which are of medical significance. The top right frame indicates which regions were considered significant. These results compare favorably with sementations performed by doctors.




Jeremy S. De Bonet
jsd@debonet.com
return to main page

Page loaded on October 23, 2017 at 06:21 PM.
Page last modified on 2006-05-27
Copyright © 1997-2017, Jeremy S. de Bonet. All rights reserved.