Synergetic relaxation labelling algorithm for segmentation of SPECT images using a connective model

A new constraint satisfaction synergetic network for 3D SPECT image segmentation using a connective model is proposed. The idea of the algorithm is based on the interaction of local synergetic dynamic processes of the system nodes with the global relaxation dynamics of the whole system. Within the approach the labels are defined on the basis of a connective model. Statistic parameters for characterizing each image class and spatial image information are proposed for solving the constraint satisfaction problem. The initialization model of the network allows a label-number-independent dimensioning of the network with n² nodes. The segmentation results obtained from SPECT images are very encouraging and show that the network is a very promising technique for image segmentation