Theoretical Bound-Guided Hierarchical VAE for Neural Image Codecs

arXiv:2403.18535v1 Announce Type: cross
Abstract: Recent studies reveal a significant theoretical link between variational autoencoders (VAEs) and rate-distortion theory, notably in utilizing VAEs to estimate the theoretical upper bound of the information rate-distortion function of images. Such estimated theoretical bounds substantially exceed the performance of existing neural image codecs (NICs). To narrow this gap, we propose a theoretical bound-guided hierarchical VAE (BG-VAE) for NIC. The proposed BG-VAE leverages the theoretical bound to guide the NIC model towards enhanced performance. We …

abstract arxiv autoencoders cs.lg eess.iv function hierarchical image images information narrow performance rate studies the information theory type vae variational autoencoders