Abstract
In signal and image processing, real data is often built by signals from multiple sources. A crucial task is thus to decompose the data into individual components, ideally the same components that originally formed the data.
Good data decomposition methods can aid in classification, recognition and understanding the data. However, we obviously have to choose a compromise between good precision and low computational costs of the algorithm.
In this paper, we present a survey of existing approaches and techniques applicable to the problem of decomposing image and other data, together with ways in which they can be effectively used.