Generalized projection algorithms with applications to optics and signal restoration

摘要

The projections-onto-constraint-sets algorithm is a powerful tool with applications in many areas, such as image reconstruction, constrained deconvolution, optimization and pattern recognition. In practice, however, it happens that its implementation is made difficult by complicated projections. In this paper we introduce an extended algorithm to mitigate these difficulties. Conditions for the exact and approximate applicability of the new method are evaluated and examples from pattern recognition and constrained deconvolution are presented.