Incorporating Ghost Module into RCAN for Super-Resolution of Satellite Images
Hiromu Ikeda, Guangxu Li, Tohru Kamiya · 2021 21st International Conference on Control, Automation and Systems (ICCAS) · 2021
With the explosion of amount of low cost satellites, satellite images have been widely used for many non-military applications, such as agriculture, landscape, and recognition of environment. Improving the image resolution to mine useful information becomes one of the immediate problems. Therefore, it is expected to improve the recognition accuracy by increasing the resolution of satellite images. Recently, deep learning technique has been proposed to increase the resolution of images. However it requires a large number of learning parameters, which results in huge computational cost. To overcome this problem, we develop a new deep learning model based on ghost module to reduce the parameters while maintaining the quality of results. We utilized Google Earth Pro satellite imagery for the network training and testing. Comparing to the classical convolutional neural network module based methods, the number of parameters used in our model was reduced 49.31 % but keeping the same level of Peak Signal-to- Noise Ratio (24.1578) and Structural Similarity (0.7174).