Optimization and Assessment of Wavelet Packet Decompositions with Evolutionary Computation

Abstract

In image compression, the wavelet transformation is a state-of-the-art component. Recently, wavelet packet decomposition has received quite an interest. A popular approach for wavelet packet decomposition is the near-best-basis algorithm using nonadditive cost functions. In contrast to additive cost functions, the wavelet packet decomposition of the near-best-basis algorithm is only suboptimal. We apply methods from the field of evolutionary computation (EC) to test the quality of the near-best-basis results. We observe a phenomenon: the results of the near-best-basis algorithm are inferior in terms of cost-function optimization but are superior in terms of rate/distortion performance compared to EC methods.

Evolutionary Optimisation of JPEG2000 Part 2 Wavelet Packet Structures for Polar Iris Image Compression

“…However, common wpb selection schemes rely on the independent evaluation of cost functions on single wpb subbands, which is not possible when recognition performance of a certain wpb has to be assessed. In earlier work [11], we have used genetic algorithms to assess the degree of optimality and to further optimize wpb subband structues. This approach is adopted for the present study where the fitness function of the evolutionary approach rating a single wpb is set to be a parameter describing recognition performance after compressing the data to JPEG2000 format using the corresponding wpb, i.e.…”

“…A key issue to apply the generic approach to the wpb subband structure selection task is to find a suitable representation of the wpb and to adapt genetic operators to the wpb tree structures [11]. The wpb can also be considered as a quadtree which needs to be transformed into a "flat" representation: in adopting principles of the heap sort algorithm, a string b of finite length L over a binary alphabet {0, 1} is used.…”

“…Classical one-or two-point crossover operators cannot be applied since in general, the resulting bitstrings do not correspond to valid wpb. Therefore, sub-tree based crossover has been introduced [11] which exchanges sub-trees and maintains a valid tree structure. Other genetic operators like mutation and selection can be applied in a standard manner.…”

Best wavelet packet bases in a JPEG2000 rate-distortion sense: The impact of header data

“…An extension to this approach employing non-additive cost functions has been developed soon after [2]. Genetic algorithms have been used [3] to assess the degree of optimality and to further optimize the subband structures found by these algorithms in earlier work. The employment of rate-distortion optimization criteria for WPB selection has been first demonstrated for classical wavelet-based compression schemes [4].…”

JPEG2000 Part 2 wavelet packet subband structures in fingerprint recognition

“…We have used genetic algorithms 12 to assess the degree of optimality and to further optimise the subband structures found by these algorithms in earlier work. The employment of rate-distortion optimisation criteria for WP subband structure selection has been first demonstrated for classical wavelet image coding schemes, 13 but has been extended later even to WP zero-tree based compression algorithms.…”