EGIC: Enhanced Low-Bit-Rate Generative Image Compression Guided by Semantic Segmentation

arXiv:2309.03244v2 Announce Type: replace-cross
Abstract: We introduce EGIC, an enhanced generative image compression method that allows traversing the distortion-perception curve efficiently from a single model. EGIC is based on two novel building blocks: i) OASIS-C, a conditional pre-trained semantic segmentation-guided discriminator, which provides both spatially and semantically-aware gradient feedback to the generator, conditioned on the latent image distribution, and ii) Output Residual Prediction (ORP), a retrofit solution for multi-realism image compression that allows control over the synthesis process by adjusting …

abstract arxiv building compression cs.cv cs.lg eess.iv feedback generative gradient image low novel oasis perception rate segmentation semantic type