An Improved Binary Tomography Reconstruction

In this paper, the binary tomographic reconstruction problem is considered. Binary tomography aims to reconstruct binary images from their projections. Possible applications can be the field of human X-ray angiography, where the aim is to reconstruct images representing blood vessels or heart chambers, using X-ray tomography methods. Injecting a contrast agent with high linear attenuation coefficient into the part of the body being examined and seek for the presence or absence of the contrast agent in certain positions. Other common applications of this field are electron tomography and industrial non-destructive testing. To reconstruct a binary image from their projections, an improved algebraic approach is proposed in this paper. An energy-minimization reconstruction model is used; this model employs a loss function to be minimized. This loss function combines three features that can be extracted from the given projections. First, data fitting term based on the given projections; second, the first two image moments that are extracted from the given projections; and third, a term that enforces the binary solution. The first two terms are expressed in terms of a linear system and the third is expressed as anon linear cost function. The projected gradient descent algorithm is then employed for this minimization process. Experimental evaluations show that reasonable results can be obtained from minimal number of projections.