Two-Dimensional Image Reconstruction from Small Amount of Projection Paths by Using Joint Two-Grade Neural Network

The first grade neural network maps the pixels to their attenuation coefficient,then the second grade maps the attenuation coefficient matrix to projection data.Finally the image is reconstructed from attenuation coefficient matrix through the trained neural network in the first grade.Such an approach can support the nondestructive test or reverse engineering for rocket engine or other huge volume products.