01533002

j

01533002 Hanke, Martin Nagy, James G. Vogel, Curtis Quasi-Newton approach to nonnegative image restorations. Linear Algebra Appl. 316, No.1-3, 223-236 (2000). 2000 Elsevier Science Inc. (North-Holland), New York, NY EN nonnegative image restorations conjugate gradient method numerical experiments constrained least squares maximum likelihood maximum entropy quasi-Newton method algorithms

doi:10.1016/S0024-3795(00)00116-6

Efficient implementations for three nonnegatively constrained image restoration schemes are introduced: constrained least squares, maximum likelihood, and maximum entropy. With a certain parametrization and using a quasi-Newton method, these algorithms become very similar. Numerical experiments show that the new approach is superior to the popular expectation maximizing method with regard to both accuracy and efficiency. F.Szidarovszky (Tucson)