Keywords: Image Compression, Image Restoration, Optimal Transport, Deep Learning
Abstract: We study an extension of lossy compression where the reconstruction distribution is different from the source distribution in order to account for distributional shift due to processing. We formulate this as a generalization of optimal transport with an entropy bottleneck to account for the rate constraint due to compression. We provide expressions for the tradeoff between compression rate and the achievable distortion with and without shared common randomness between the encoder and decoder. We study the examples of binary, uniform and Gaussian sources (in an asymptotic setting) in detail and demonstrate that shared randomness can strictly improve the tradeoff. For the case without common randomness and squared-Euclidean distortion, we show that the optimal solution partially decouples into the problem of optimal compression and transport and also characterize the penalty associated with fully decoupling them. We provide experimental results by training deep learning end-to-end compression systems for performing denoising on SVHN and super-resolution on MNIST suggesting consistency with our theoretical results.
One-sentence Summary: We consider the novel task of cross-distribution lossy compression and characterize it as an optimal transport problem under an entropy constraint, then provide experimental results to demonstrate the principles suggested by our theory.