Guest Editorial: Special Issue on Blind Source Separation and De-convolution in Imaging and Image Processing

摘要

The relatively recent development of robust methods for blind-source separation (BSS) provides a new algorithmic toolset for addressing under-determined ill-posed problems in imaging, vision and image processing. In BSS a set of observations are assumed to be composed of a mixture (often assumed linear) of some so-called sources, in which neither the sources nor the mixing coefficients are known to the observer. The goal is to "blindly" recover the sources, either with or without recovering the mixing matrix, given the set of observations. Many approaches to BSS have been developed, employing probabilistic information, theoretic, geometric or varia-tional-based techniques. BSS methodologies and algorithms have already shown great promise in imaging sciences, since the linear mixing model, or the approximations and variations of it are often encountered in many aspects of imaging, image processing, and analysis—e.g., unmixing of tissue signatures by means of MRI or optical imaging, spectral unmixing, linear data-adaptive feature