Abstract. We apply independent component analysis (ICA) for learning an efficient color image representation of natural scenes. In the spectra of single pixels, the algorithm was a...
Te-Won Lee, Thomas Wachtler, Terrence J. Sejnowski
Statistical methods, such as independent component analysis, have been successful in learning local low-level features from natural image data. Here we extend these methods for le...
In this paper, linear multilayer ICA (LMICA) is proposed for extracting independent components from quite high-dimensional observed signals such as large-size natural scenes. Ther...
The problem of blindly separating signal mixtures with fewer mixture components than independent signal sources is mathematically ill-defined, and requires suitable prior informat...