Study On Optimization Theory And Algorithms Of Stack Filters

Stack filters are a class of slip window nonlinear digital filters that satisfy a weak superposition property known as the threshold decomposition and stacking property. They are suitable to parallel processing and VLSI realization. Stack filters include many kinds of nonlinear filters in theory. Thus, the study of stack filters has important theory value and practice meaning.The key problem of the design of stack filters is optimization, so this thesis is dedicated to the search on optimization algorithms of stack filters. The main contents and contributions of this thesis are showed as follows:The optimization of stack filters usually uses genetic algorithm, this paper presents optimizing stack filters using bionic intelligence algorithms under the MAE criterion. 1. Realize the optimal stack filters by particle swarm algorithms, and it can receive better effect and run faster than using genetic algorithms. 2. On the base of basic clone selection algorithms, presents an improved algorithm. The performance of stack filters can be advanced more on effect and speed using this method. 3. Use the fast constringency characteristic of the ant colony algorithms, amend its operator. Then taking its result as the initial value, combine the improved clone selection algorithm to advance the optimized speed much more.This thesis studies the optimization problem of stack filters based on pth-order error criterion using global optimization algorithms. And simulates the optimal stack filters based on 3th-order error norm and 4th-order error norm separately. The results show that they have better vision effect than that optimized under MAE norm. And their capability of suppressing noises is greater than that optimized under MAE criterion or under MSE criterion.On the base of NWMAE criterion, this thesis presents an adaptive power coefficient NWMAE criterion. The power coefficient of this method can be auto regulate according to the signal value of current window. The optimal stack filters received under this norm can keep details well while suppress noises efficiently.