Structural optimization of wavelet packets using swarm algorithms

Image compression is an essential task since data amount increases by technological developments. Although big size data may be helpful while retrieving information, storing, processing, transmission etc., is expensive. Therefore, image compression techniques are introduced to represent the data by less bits. One of the compression techniques, wavelet transform, is used especially to compress images. By the wavelet packets decomposition, both approximation and detail coefficients of an image are extracted repeatedly up to a filtering level. In order to decompose an image by wavelet packets, there are some parameters to be set such as main wavelet type, filtering level, and threshold values at each level. Selecting the best values for these parameters affects the performance of the compression. Therefore, assigning the parameters that yield the optimum compression is a design problem. Moreover, since the filtering type and filtering level are related with the topology of the filter and the level of filtering changes the number of parameters to be optimized, the problem can be considered as a structural optimization problem. In order to solve this problem, two swarm-intelligence based optimization algorithms, Particle Swarm Optimization and Artificial Bee Colony algorithms, are employed and compared in terms of compression and quality metrics.