Abstract—This paper proposes a dual-energy computed tomography (DECT) image super-resolution (SR) reconstruction scheme. The SR reconstruction scheme is based on sparse representation theory and dictionary learning of low-resolution and high-resolution image block pairs to improve the poor quality of low-resolution dual-energy CT images obtained in clinical practice. The proposed strategy is based on the idea of sparse representation, that is, image blocks can be well represented by sparse coding elements from over-complete dictionaries. We have jointly trained two pairs of dictionaries for high spectral CT images and low spectral CT images, and each pair of dictionaries contains dictionaries of low-resolution and high-resolution image blocks. Low-resolution dual-energy CT images can be represented by a low-resolution dictionary trained on high spectral CT images and a dictionary trained on low spectral CT images multiplied by sparse representation coefficients. And the sparse representation coefficient is multiplied by the corresponding high-resolution dictionary to reconstruct the high-resolution CT images. With an appropriate amount of iterative operations, the reconstructed high-resolution image can obtain better image quality and clearer visual effects. Experiments prove that the dual-energy CT image reconstructed by the sparse representation method has improved peak signal-to-noise ratio and structural similarity, and the image details and textures are clearer.

super-resolution,dual-dictionary learning,sparse representation,dual-energy computed tomography