@MASTERSTHESIS\{IMM2014-06840,
    author       = "K. K. Nielsen",
    title        = "Block Algebraic Methods for {3D} Image Reconstructions on GPUs",
    year         = "2014",
    school       = "Technical University of Denmark, Department of Applied Mathematics and Computer Science",
    address      = "Richard Petersens Plads, Building 324, {DK-}2800 Kgs. Lyngby, Denmark, compute@compute.dtu.dk",
    type         = "",
    note         = "{DTU} supervisor: Per Christian Hansen, pcha@dtu.dk, {DTU} Compute",
    url          = "http://www.compute.dtu.dk/English.aspx",
    abstract     = "This report investigates the possibility of making effective three dimensional tomographic reconstructions on General Purpose Graphical Processing Units (GPGPU). Tomography is a technique to recreate the interior of an object, from measurements of absorbed energy from projections taken at different angles. When these calculations are done on a computer it is called Computed Tomography (CT).
This will often lead to a underdetermined system of equations which therefore is highly influenced by noise. The work described in this report will focus on the implementation of a family of methods which contains a natural regularisation in terms of the number of used iterations.
As a starting point there was used an open source package called {ASTRA} which contains multiple methods for doing tomographic reconstructions on {GPGPU}’s. One of these methods (SIRT) can be considered as a special case of the family of methods this report is focusing on. Using the methods described in this report it is possible to make faster tomographic reconstructions than by using the similar method from {ASTRA}.
The implementation was made in {CUDA} C and C++ and contains wrappers which exposes the code to Matlab."
}