CMMR'10 Proceedings of the 7th international conference on Exploring music contents
Algorithms for nonnegative matrix factorization with the β-divergence
Neural Computation
Hi-index | 0.01 |
We present a multiplicative algorithm for image reconstruction, together with a partial convergence proof. The iterative scheme aims to maximize cross Burg entropy between modeled and measured data. Its application to infrared astronomical satellite (IRAS) data shows reduced ringing around point sources, compared to the EM (Richardson-Lucy) algorithm