Application of SVD by Means of ICA
Valeriu D. Vrabie, Jérôme Igor Mars · 2002
P027 APPLICATION OF SVD BY MEANS OF ICA Abstract 1 The Singular Value Decomposition (SVD) is used to perform a separation of the initial seismic dataset into two or more complementary subspaces. This decomposition provides two orthogonal matrices made up of normalized wavelets and propagation vectors. The constraint of orthogonality imposed for the propagation vectors implies errors in the estimated subspaces because it forces the normalized wavelets to be a mixture of source waves. We propose in this paper a modified SVD based Independent Component Analysis (ICA) decomposition in order to relax the non-physically justified orthogonality of the propagation vectors