Abstract
The usual quadratic programming approach for image reconstruction is reformulated by the introduction of certain inequality constraints. These constraints constitute additional informations about a problem. We shall concern ourselves with linear inequality constraints only. A variety of methods have been presented in the literatures. We mention particularly one algorithm that has attracted serious attention to the numerical stability of the method and which has been basically derived from the conjugate gradient method. The suitability of the method is tested by the reconstruction of a human head from simulated X-ray data and reconstruction of a gray level picture digitized on a 128 × 128 grid.
Indexing terms: