It uses an initial algorithm developed by Antonin Chambolle described in An Algorithm for Total Variation Minimization and Applications. I added the link to the Reconstruction codes section of the big picture.
Thank you Gabriel.
Tuesday, August 19, 2008
CS: Chambolle's Algorithm for the Resolution of Compressed Sensing with TV regularization
When reading this Future of Science entry, I was wondering if I knew of examples where the web 2.0 was being part of the scientific process as opposed to just reporting on it. Yesterday, Laurent Jacques wrote an entry on the potential ability for SPGL1 to perform TV minimization in a blog entry entitled SPGL1 and TV minimization ?. Gabriel Peyre had some thoughts on the matter as can be seen in the comments section and he went on compiling a very nice explanation and an attendant commented code in this page: Chambolle's algorithm for the resolution of compressed sensing with TV regularization. Gabriel thinks "it’s like the fastest algorithm on earth to solve CS-TV"[ see the caveat in the comment section of this entry! ]
It uses an initial algorithm developed by Antonin Chambolle described in An Algorithm for Total Variation Minimization and Applications. I added the link to the Reconstruction codes section of the big picture.
Thank you Gabriel.
It uses an initial algorithm developed by Antonin Chambolle described in An Algorithm for Total Variation Minimization and Applications. I added the link to the Reconstruction codes section of the big picture.
Thank you Gabriel.
Just to make things clear: I do not really think this iterative algorithm is the fastest one. It was more like a provocation because of so many articles claiming having the fastest code on earth to solve L1 or TV ... At least this one is simple to implement !
ReplyDeleteTo quote Chambolle's paper "However, we found out that our method is quite slow, and does not seem to be a great improvement with respect to standard methods". As usual, Antonin is very pesimistic !
The paper of Antonin can be found here.