Generalized Principal Component Pursuit

min nuclear_norm(L) + beta*||W(S)||_1 subject to ||y-F(S+L)|_2 < err

Description

This is a generalized version of Principal Component Pursuit (PCP) where the sparsity is assumed in a transform domain and not in measurement domain. Moreover the samples obtained are lower dimensional projections.
% Inputs
% y - observation (lower dimensional projections)
% F - projection from signal domain to observation domain
% W - transform where the signal is sparse
% beta - term balancing sparsity and rank deficiency

% Outputs
% S - sparse component
% L - low rank component

requires sparco for defining operators
http://www.cs.ubc.ca/labs/scl/sparco/

• MATLAB release: MATLAB 7.9 (2009b)
• Other requirements: Requires Sparco http://www.cs.ubc.ca/labs/scl/sparco/

[Zhang Xuande]

great!!