https://hal-univ-pau.archives-ouvertes.fr/hal-01912872Luthon, FranckFranckLuthonLIS - Laboratoire des images et des signaux - UJF - Université Joseph Fourier - Grenoble 1 - INPG - Institut National Polytechnique de Grenoble - CNRS - Centre National de la Recherche ScientifiquePopescu, George V.George V.PopescuCaplier, AliceAliceCaplierLIS - Laboratoire des images et des signaux - UJF - Université Joseph Fourier - Grenoble 1 - INPG - Institut National Polytechnique de Grenoble - CNRS - Centre National de la Recherche ScientifiqueAn MRF Based Motion Detection Algorithm Implemented on Analog Resistive NetworkHAL CCSD1994[INFO] Computer Science [cs][INFO.INFO-TS] Computer Science [cs]/Signal and Image ProcessingLANNUZEL, GAELLEEklundh, Jan-Olof2018-11-05 17:18:032022-06-26 09:14:082018-11-05 17:18:03enConference papers10.1007/3-540-57956-7_191We present an algorithm based on MRF modelling for motion detection in image sequences and give a modified version for implementation on analog resistive network. Energy minimization is realized by a network relaxing to its state of minimal power dissipation. It takes a few nanoseconds and replaces advantageously time consuming stochastic or suboptimal deterministic relaxation algorithms. The elementary cell of the network is presented along with the environment needed to feed it with the required inputs. Two network architectures are proposed, derived from CCD camera principle. Software simulations of a 128\texttimes128 network demonstrate the good behaviour of the modified algorithm on real sequences. Electrical simulations of a 16\texttimes16 network with ideal components give promising results. Implementation of the CMOS circuit with VLSI technology is under study at our laboratory. \textcopyright Springer-Verlag Berlin Heidelberg 1994.