Chair: Ken Sauer, University of Notre Dame, USA
Kaaren May, Imperial College of Science Technology and Medicine (U.K.)
Tania Stathaki, Imperial College of Science Technology and Medicine (U.K.)
Anthony Constantinides, Imperial College of Science Technology and Medicine (U.K.)
A Simulated Annealing Genetic Algorithm (SAGA) is presented for blind restoration of a nonlinearly-degraded image with additive Gaussian noise. The degradation is modelled by a quadratic Volterra filter. Deconvolution of the original image and the unknown Volterra filter is formulated as a constrained optimisation problem, the cost function of which is minimised by the SAGA.
Ken Sauer, University of Notre Dame (U.S.A.)
Charles A. Bouman, Purdue University (U.S.A.)
The objective of this paper is to investigate a new non-iterative paradigm for image reconstruction based on the use of nonlinear back projection filters. This method, which we call nonlinear back projection (NBP), attempts to directly model the optimal inverse operator through off-line training. Potential advantages of the NBP method include the ability to better account for effects of limited quantities and quality of measurements, image cross-section properties, and forward model non-linearities. We present some preliminary numerical results to illustrate the potential advantages of this approach and to illustrate directions for future investigation.
Constantine Kotropoulos, Aristotle University of Thessaloniki (Greece)
Maria Gabrani, Drexel University (U.S.A.)
Ioannis Pitas, Aristotle University of Thessaloniki (Greece)
Adaptive multichannel L-filters based on marginal ordering axe studied in this paper when structural constraints such as the location-invariance or the unbiasedness axe imposed on the filter coefficients. Two novel adaptive algorithms are derived by using Frost's algorithm for minimizing the Mean Squared Error subject to the above-mentioned constraints in the LMS and in the LMS-Newton algorithms. It is demonstrated by experiments that the Frost-LMS algorithm has a faster convergence rate than the Frost LMS-Newton algorithm but it yields a higher steady-state MSE than that.
Fathi M. Salam, Michigan State University (U.S.A.)
Gamze Erten, Innovative Computing Technologies Inc. (U.S.A.)
This work bridges the gap between activities motivated from statistical signal processing, neuromorphic systems, and microelectronic implementation techniques for blind separation and recovery of mixed signals. The composition adopts both discrete-time and continuous-time formulations with a view towards implementations in the digital as well as the analog domains of microelectronic circuits. This paper focuses on the development and formulation of dynamic architectures with adaptive update laws for multi-source blind signal separation/recovery.
Mounir Sayadi, University of Tunis (Tunisia)
Farhat Fnaiech, University of Tunis (Tunisia)
Daniel Bastard, Equipe Signal & Image ENSERB (France)
In this paper, a fast approach for training the adaptive neural stack filter based on the Recursive Least Square algorithm is presented. The architecture of the neural stack filter is transformed in order to allow the fast recursive estimation of the adaptive neural filter weights. This approach is compared to the Least Mean Square training algorithm (also called Back-propagation algorithm in case of multilayer neural network). Computer simulations on restoration of noisy images are presented. It is shown that the proposed fast training algorithm is 2.5 time faster than the LMS algorithm.