Optimization design of orthogonal filter banks for image coding via multi-objective genetic algorithm

In this paper, we propose an optimisation method based on a multi-objective Genetic Algorithm (GA) for the design of orthogonal filter banks in an image coding scheme. A parameterization is used to achieve perfect reconstruction orthonormal FIR filter banks with a first order regularity. We search the optimal parameter set according to the coding gain, frequency selectivity and the group delay characteristics. Particularly, to design near linear phase filter banks, the group delay flatness in the filter pass-band is introduced as an objective to be minimised. We formulate the optimization problem as multi-objective and we use the Non-dominated Sorting Genetic Algorithm approach (NSGAII) to solve this problem by searching solutions that achieve the best compromise between the different objectives criteria, these solutions are known as Pareto Optimal Solutions. From experimental results, our new optimal filter banks are shown to outperform significantly the Daubechies orthogonal filter banks for the majority of test images.