THE MULTI-SENSOR IMAGE SEGMENTATION DATA SET
of input (from the ImageFusion.org and other collections) and fused images which were segmented by hand (kindly provided by Mr John Lewis and Dr Stavri Nikolov, University of Bristol, UK, and by Dr Lex Toet, TNO, The Netherlands)
 (To download a full-size image click on the thumbnail image.)

General information:

These are the results of human segmentations of a number of images on the 3rd and 5th May 2005 produced at the University of Bristol, UK. The aim was to produce a multi-sensor segmentation data set similar to the Berkeley Segmentation Data Set. A version of the image segmentation tool provided on their web site was used in our experiment.

Each user was shown 5 images randomly selected for all images but based on certain rules: (a) Each subject sees only one image from each visible/ IR/fused set (i.e. either the visible, IR or a fused image); (b) Each subject should segment at least one visible image, one IR image and one fused image; and, (c) An image should not be allocated for a second time until all available images have been allocated once; and not for a third time until all available images have been allocated twice; etc. In addition, all the images were segmented by hand after the experiment by Dr Lex Toet and these results are provided as well .

Naming conventions:

  • original file name_user_im.jpg - original image
  • original file name_user_im_map.png - a numbered map showing segmentation regions from 0..N where N is the number of regions (these files are not shown on the web pages below but are included in the zip files)
  • original file name_user_im_olay.png - an overlay of the segmentation on the original image.
  • original file name_user_im.seg - the segmentation file produced by Berkeley segmentation program. More information about the segmentation format can be found here. The files segmented by Dr Toet have no .seg files accompanying them and have not been included in the all users' segmentation overlays in the web pages below.

    •  Hand segmented multi-sensor input and fused images:

  • TNO UN Camp (frame 1811): visible, infrared, pyramid fused, DWT fused, CWT fused, all in one zip file (2584KB)
  • TNO Trees (frame 4906): visible, infrared, pyramid fused, DWT fused, CWT fused, all in one zip file (2238KB)
  • TNO Trees (frame 4917): visible, infrared, pyramid fused, DWT fused, CWT fused, all in one zip file (2150KB)
  • TNO Dune (frame 7404): visible, infrared, pyramid fused, DWT fused, CWT fused, all in one zip file (2126KB)
  • TNO Kayak (frame 7118a): visible, infrared 1, infrared 2, pyramid fused, DWT fused, CWT fused, all in one zip file (7474KB)
  • TNO Kayak (frame 7436a): visible, infrared 1, infrared 2, pyramid fused, DWT fused, CWT fused, all in one zip file (4805KB)
  • TNO Kayak (frame e518a): visible, infrared 1, infrared 2, pyramid fused, DWT fused, CWT fused, all in one zip file (6588KB)
  • Octec (frame 2): visible, infrared, pyramid fused, DWT fused, CWT fused, all in one zip file (5719KB)
  • Octec (frame 21): visible, infrared, pyramid fused, DWT fused, CWT fused, all in one zip file (5702KB)
  • Equinox Face (frame 2374-9-1-0): visible, infrared, pyramid fused, DWT fused, CWT fused, all in one zip file (2008KB)
  • Bristol Queen's Road: visible, infrared, pyramid fused, DWT fused, CWT fused, all in one zip file (5923KB)
    • Acknowledgements:

    This work has been partially funded by the UK MOD Data and Information Fusion Defence Technology Centre (DIF DTC). The original “UN Camp”, “Trees”, “Dune” and “Kayak” visible and IR images are kindly supplied by TNO, The Netherlands. The "Octec" images are provided by David Dwyer of Octec Ltd. The “Face” images are taken from the Human Identification at a Distance (HID) Data Set, produced by Equinox Corp. All intellectual rights for the input images are with the suppliers of these individual images (see the respective web pages).

    You are free to download the full Multi-Sensor Segmentation Data Set or a part of it for non-commercial research and educational purposes. Work based on the data set should cite the paper below and acknowledge the suppliers of the individual images.

    References:

  • J. J. Lewis, S. G. Nikolov, A. Toet, D. R. Bull, C. N. Canagarajah, Uni-modal versus joint segmentation for region-based image fusion, Proceedings of the 9th International Conference on Information Fusion, (Fusion 2006), Florence (Italy), 10-13 July 2006