Name
Abstract | ||
|---|---|---|
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. |
Year | DOI | Venue |
|---|---|---|
2019 | 10.1109/ICCV.2019.00031 | 2019 IEEE/CVF International Conference on Computer Vision (ICCV) |
Keywords | DocType | Volume |
decoded image,generative adversarial networks,extreme learned image compression,learned image compression system,extremely low bitrates,multiscale discriminator,generative learned compression objective,visually pleasing results,previous methods fail,semantic label map | Conference | 2019 |
Issue | ISSN | ISBN |
1 | 1550-5499 | 978-1-7281-4804-5 |
Citations | PageRank | References |
8 | 0.53 | 4 |
Authors | ||
5 | ||
Name | Order | Citations | PageRank |
|---|---|---|---|
| Eirikur Agustsson | 1 | 257 | 13.89 |
| Michael Tschannen | 2 | 143 | 13.58 |
| Fabian Mentzer | 3 | 60 | 5.08 |
| Radu Timofte | 4 | 1880 | 118.45 |
| Luc Van Gool | 5 | 27566 | 1819.51 |