Generative Adversarial Networks for Extreme Learned Image Compression
This addresses the problem of efficient image storage and transmission for applications like mobile or bandwidth-limited environments, though it is incremental as it builds on existing GAN and compression techniques.
The paper tackles extreme low-bitrate image compression by using a GAN-based framework that synthesizes missing details, achieving visually pleasing results where prior methods fail with artifacts, and a user study shows it outperforms state-of-the-art methods even with less than half the bits.
We present a learned image compression system based on GANs, operating at extremely low bitrates. Our proposed framework combines an encoder, decoder/generator and a multi-scale discriminator, which we train jointly for a generative learned compression objective. The model synthesizes details it cannot afford to store, obtaining visually pleasing results at bitrates where previous methods fail and show strong artifacts. Furthermore, if a semantic label map of the original image is available, our method can fully synthesize unimportant regions in the decoded image such as streets and trees from the label map, proportionally reducing the storage cost. A user study confirms that for low bitrates, our approach is preferred to state-of-the-art methods, even when they use more than double the bits.