Parallel Implementation of Tomographic Reconstruction Algorithms on Bus-Based Extended Hypercube

The convolution backprojection (CBP) and Fourier inversion method (FIM) are popular tomographic image reconstruction techniques. However, the time required to reconstruct an image has been a major drawbacks associated with these techniques. We have designed a bus-based extended hypercube (BEH) that combines the positive features of both the bus-based systems and the hypercube. We have executed the CBP and FIM algorithms on a hypercube, extended hypercube and on BEH. The reasons for high speedup of the BEH for image reconstruction tasks could be attributed to fast collective communication and collective computation algorithms of BEH.