Laplacian-guided Entropy Model in Neural Codec with Blur-dissipated Synthesis

arXiv:2403.16258v1 Announce Type: cross
Abstract: While replacing Gaussian decoders with a conditional diffusion model enhances the perceptual quality of reconstructions in neural image compression, their lack of inductive bias for image data restricts their ability to achieve state-of-the-art perceptual levels. To address this limitation, we adopt a non-isotropic diffusion model at the decoder side. This model imposes an inductive bias aimed at distinguishing between frequency contents, thereby facilitating the generation of high-quality images. Moreover, our framework is equipped with a …

abstract art arxiv bias codec compression cs.cv cs.it cs.lg data diffusion diffusion model eess.iv entropy image image data inductive math.it quality state synthesis type