Title :
Autonomous clustering of fractal patterns via Hausdorff potentials
Author_Institution :
Fac. of Eng., Osaka Inst. of Technol., Japan
Abstract :
A nondeterministic kinetics is introduced in image plane for autonomous clustering of fractal attractors associated with contraction mappings. By reformulating 2D Newton potential in terms of the Hausdorff distance, both the attribution to fixed points of contraction mappings and the consistency of fixed point estimates are evaluated. Attracted by fixed point estimates, feature points are aggregated to successively organize discrete clusters structurally consistent with the mapping set. The proposed scheme was implemented and verified through simulation studies
Keywords :
feature extraction; fractals; image matching; stochastic processes; 2D Newton potential; Hausdorff potentials; autonomous clustering; contraction mappings; fixed point estimates; fractal attractors; fractal patterns; nearest neighbour aggregation; pattern clustering; self similarity; Data mining; Electrostatics; Fractals; Gravity; Image segmentation; Kinetic theory; Large-scale systems; Pattern clustering; Stochastic processes; Visual perception;
Conference_Titel :
SICE Annual, 1999. 38th Annual Conference Proceedings of the
Conference_Location :
Morioka
Print_ISBN :
4-907764-13-8
DOI :
10.1109/SICE.1999.788701
