Inversion of noisy Radon transform by SVD based needlets (Q1045707)

In this exciting paper, the authors considered the problem of inverting noisy observations of the Radon transform. They observed that the singular value decomposition (SVD) basis is very stable for inversion but not quite capable of representating local features of images. Therefore the authors introduced an estimation method based on localized frames based on orthogonal polynomials on the ball, and investigated the possible losses in the minimax framework. Simulations show that the new estimation method has much better performance than the classical SVD method.