An Adaptive High-Fidelity Image Compression Framework for Internet of Vehicles

This paper proposes a new adaptive high-fidelity image compression solution to achieve a high compression ratio with the least distortion using a generative adversarial network. This work focuses on preserving the details by compressing the salient regions in the image with a high bit rate to guarantee the generation of high-quality outputs that sustain most of its characteristics. The image background is compressed with a lower bit rate. This work is tested against the Kodak, CLIC, MOTS, and UADTV datasets based on the bit-per-pixel rate where the results prove that our work achieves the highest quality with the lowest rate. To achieve a lower bit rate, the arithmetic coding algorithm is applied to the compression sequence which reduces the rate by 35%. With the achieved low bit rate, our work boosts the rate of image transmission by a factor of more than 2.