The Formosat-8 image super-resolution system is to enhance the Formosat-8 image from 1 meter resolution to sub-meter resolution. The super-resolution algorithm design based on sub-pixel shift linear panchromatic sensor pairaccelerating processing system for linear sensor. In addition, the super-resolution algorithm develops based on deep learning network to obtain high-resolution images for frame sensor.

The optical system of Formosat-8 satellite is different from the traditional imaging system. Formosat-8 satellite has sub-pixel shift linear sensor pair,the traditional imaging system is frame sensor. Therefore, the concept of super resolution algorithm is different between Formosat-8 satellitetraditional imaging system. In addition, the frame sensor of Formosat-8 satellite will face the problems of geometric deformation, relief displacementresolution change. In the deep learning, the Convolutional Neural Networktraining dataset should be designedconstructed for overcoming these problems.

The image super-resolution technique can be widely applied in various fields, such as satellite image processing, medical image processing,the multimedia industry. In satellite image processing, the super-resolution technique can be used to generate rich geographic information. In medical image processing, the super-resolution technique can be used in CT, MRT,X-ray images, high-resolution images are helpful for doctors to make a diagnosis. In the multimedia industry, the super-resolution technique can be used to generate high-resolution picturescreate a comfortable viewing experience.