Super-resolution Method for Radiation Image Based on Deep Learning

In the security check system, the spatial resolution of radiation image generated by digital radiography is often so low that reduces the image quality. In this work, a super-resolution method based on deep learning was proposed. Using the convolution neural network with residual block, the method trained the radiation image sample in dataset and found the mapping function of low-resolution image to high-resolution image. The experiment result shows that the super-resolution method can deliver superior performance compared with other traditional methods while maintaining an excellent speed. The study result indicates the great potential of deep learning in radiation image processing.