Detection of concealed objects, such as explosives or illegal drugs, using X-rays is confounded when the object is formed into a thin sheet. We have explored the potential of binary image restoration to detect such thin objects in two- and three-dimensional images. The use of a weighted mean-square error estimate to perform the restoration optimizes the restoration to place emphasis on the infrequent, but significant local structure of the thin objects. Experimental results show the restoration of thin lines and curves in two-dimensional data, and thin sheets in three-dimensional tomographic data.

Contents

To Contact the Authors:

• Jeff Boyd is at the University of Calgary, Department of Computer Science. Send e-mail to:
  boyd@cpsc.ucalgary.ca

• Jean Meloche is at the University of British Columbia, Department of Statistics. Send e-mail to:
  jean@stat.ubc.ca