Evolvable Hardware as Non-linear Predictor for Image Compression

Evolvable hardware (EHW) is a new technology, which was discovered at intersection of artificial intelligence and the circuit design. Unique hardware architecture is searched for each task. The circuit connection is subject of evolution and the function of such circuit can adapt to dynamic changing environment. Genetic algorithm is used to simulate evolution – its chromosomes encode potential solutions (configuration bits of used reconfigurable chip). Simply, we can speak about hardware evolution. A concept for modelling of EHW and EHW-based applications is explained. Models of reconfigurable circuit and evolution are used in the application – lossy image compression – where EHW works as non-linear predictor for block of data. The new approach to compression quality control is described too. The results are compared with JPEG algorithm.