Spectral CT reconstruction using image sparsity and spectral correlation

Spectral data across energy bins are highly correlated but the signal to noise ratio is generally poor for each energy bin due to the narrow bin width and resultant quantum noise. To address this problem, in this paper we propose a spectral CT reconstruction approach that simultaneously reconstructs x-ray attenuation coefficients in all the energy bins, aided by intra-image sparsity and inter-image similarity. In addition to a sparsifying transform in terms of total variation (TV) and spectral means (SM) are respectively utilized for similarity measurement. To facilitate the quantification of the similarity, we design a linear mapping function to minimalize image differences between different energy bins. Experimental results show that the proposed algorithms qualitatively and quantitatively outperform existing iterative algorithms for spectral CT reconstruction.