Low-dose CT image super resolution using a model-based framework with CNN prior

Improve low-dose CT (LDCT) imaging quality is a hot research issue in the medical imaging field. However, current image post-processing method mainly focus on denoising, ignoring the limitations of image resolution. To this end, a new model-based LDCT image super resolution based on CNN prior is proposed in this paper. Meanwhile, the Half Quadratic Splitting (HQS) algorithm is selected to accelerate high resolution LDCT calculation process. The convolutional neural network (CNN) prior is pre-trained and integrated into the HQS algorithm to regularize the model-based LDCT SISR problem. Experimental results demonstrate that our method can improve the resolution of low-dose CT images effectively.