| In this thesis, various methods of blind source separation by independent component analysis are described. Blind signal processing has recently received tremendous attention because of its existing and potential applications in signal and image processing such as speech recognition systems, telecommunications, medical imaging, pattern recognition, remote sensing, etc. The goal of the blind source separation is to recover independent sources from sensor observation which are unknown linear mixtures of the unobserved source signals. In contrast to correlation-based transformation such as the principal component analysis, the blind techniques not only decorrelate the signals (second-order statistics) but also reduce higher-order dependencies, attempting to make the signals as independent as possible. The thesis describes, reviews, and compares the existing algorithms with the emphasis on applications in image processing. Extensive computer simulations illustrate and confirm the usefulness and performance of the developed algorithms. |
