Terahertz Deep Learning Super Resolution Imaging Training on Sinogram

We presented a supervised deep learning method to decrease six-fold data acquisition time and to suppress terahertz image artifact levels based on sinogram from terahertz time-domain spectroscopy (THz- TDS). Our method delivered superior image quality than conventional bicubic interpolation method with 0.9612 and 0.0155 in mean square error (MSE) and structural similarity index (SSIM), respectively.