Image motion analysis using scale space approximation and simulated annealing
This paper addresses the problem of motion estimation in sequences of remotely sensed images of the sea. When the temporal sampling period is low the estimation of the velocity field can be done by finding the correspondence between structures detected in the images. The scale space aproximation of these structures using the wavelet multiressolution is presented. The correspondence is solved using a simulated annealing technique which assures the convergence to high quality solutions.
KeywordsSimulated Annealing Multiresolution Analysis High Quality Solution Hopfield Neural Network Correspondence Problem
Unable to display preview. Download preview PDF.
- 1.Aarts, E., Korst, J.: Simulated annealing and Boltzmann machines: a stochastic approach to combinatorial optimization and neural computing. John Wiley & Sons (1989)Google Scholar
- 2.Aarts, E., Van Laarhoven, P.: A new polynomial time cooling schedule. Proc. IEEE Int. Conf. on Computer Aided Design. Santa Clara (1985) 206–208.Google Scholar
- 3.Côte, S., Tatnall, A.R.L.: Estimation of ocean surface currents from satellite imagery using a Hopfield neural network. Third Thematic Conference on Remote Sensing for Marine and Coastal Environments I Seattle (1995) 538–548.Google Scholar
- 7.Ikeda, M.: Mesoscale variability revealed with sea surface temperature imaged by AVHRR on NOAA satellites. Oceanographic applications of remote sensing. CRC Press. (1995) 3–14Google Scholar
- 9.Parisi V., et al.: A Hopfield neural network to track drifting buyos in the ocean. Proc. of the Ocean's 98 Conference II (1998) 1010–1018Google Scholar
- 10.Waku, J., and Chassery, J.M.: Wavelets and multi-Scale representation of discrete boundary. Proceedings of 11th. ICPR. The Hague (1992) 680–683.Google Scholar