Enhanced Source Separation by Morphological Component Analysis

This paper describes two extensions of the recent morphological component analysis (MCA) method to multichannel data. MCA takes advantage of the sparse representation of structured data in large overcomplete dictionaries to separate features in the data based on their morphology. It was shown to be an efficient technique in such problems as separating an image into texture and piecewise smooth parts or for inpainting applications. A first extension, MMCA, achieves a similar source separation objective based on morphological diversity. A second extension, GMMCA, takes advantage of the highly sparse representations of the sources that can be built using MCA. Indeed, sparsity is now generally recognized as a valuable property for blind source separation. The efficiency of MMCA and GMMCA is confirmed in numerical experiments