UNI-MB - logo
UMNIK - logo
 
E-viri
Celotno besedilo
Recenzirano Odprti dostop
  • Enhancing Very High Resolut...
    Jangir, Sandeep Kumar; Bahmanyar, Reza

    IEEE journal of selected topics in applied earth observations and remote sensing, 12/2023
    Journal Article

    Employing a two-step pipeline that encompasses an Image-to-Image Translation (I2I) and a Super-Resolution (SR) network, we significantly enhance satellite images with a Ground Sample Distance (GSD) of 30 centimeters (cm) to a superior 15 cm GSD. Our translation network learns from the characteristics of satellite images and replicates these onto aerial images with a GSD of 15 cm, creating a tailored training dataset. The super-resolution network then uses this dataset to train and render enhanced satellite images at a GSD of 15 cm. This innovative approach can provide a cost-effective substitute for commercial high-definition (HD) images, and broadens the usage of high-quality data across various applications.