Alternating minimization algorithm based optimization method for Total Variation

Series: 
ACO Student Seminar
Friday, February 1, 2013 - 13:00
1 hour (actually 50 minutes)
Location: 
Skiles 005
,  
CSE, Georgia Tech
,  
 In this talk, we are going to introduce Linearized Proximal Alternating Minimization Algorithm and its variants for total variation based variational model. Since the proposed method does not require any special inner solver (e.g. FFT or DCT),  which is quite often required in augmented Lagrangian based approach (ADMM), it shows better performance for large scale problems. In addition, we briefly introduce new regularization method (nonconvex higher order total variation).