ARCHE: Autoregressive Residual Compression with Hyperprior and Excitation

ArXi:2603.10188v1 Announce Type: cross Recent progress in learning-based image compression has nstrated that end-to-end optimization can substantially outperform traditional codecs by jointly learning compact latent representations and probabilistic entropy models. However, many existing approaches achieve high rate-distortion efficiency at the expense of increased computational cost and limited parallelism.