Super-Resolution of Images Received from UAV
Victor M. Sineglazov, Kyrylo Lesohorskyi, Olena Chumachenko · 2023
This work concerns the problem of real-time image super-resolution (RTISR) for drone navigational video feed. A dual-step convolutional neural network approach based on residual learning and sub-pixel convolutions is proposed. The feature maps are extracted in low-dimensional space via residual feature extraction head and then passed through subpixel deconvolutional layers. The results indicate that the efficiency of the proposed approach is comparable to the state-of-the-art approach, but further improvements can be made to increase the throughput of the proposed architecture.