Proposing a novel Frequency-Aware Transformer (FAT) block for learned image compression to capture multiscale and directional frequency components efficiently.
