Region growing is a very useful technique for image segmentation. Its
efficiency mainly depends on its aggregation criterion. In the present
paper, a new algorithm is proposed with a homogeneity criterion based
on an adequate tuning between spatial neighbourhood and histogram nei
ghbourhood. It differs from other techniques by reconsidering the pixe
l (or voxel) assignments on each step by a process which minimizes var
iance through special dilations. Thus, the region created by an initia
l seed can be non-connected and possibly does not contain this seed. E
xamples are given in dental surgery for 2D X-Ray images (and their ass
ociated 3D block) and for 3D images acquired by the Morphometre, the n
ew 3D scanner constructed by GEMSE (General Electric Medical Systems).
(C) 1997 Published by Elsevier Science B.V.