ProxIQA: A Proxy Approach to Perceptual Optimization of Learned Image Compression
This addresses the issue of perceptual inconsistency in image compression for users needing efficient storage or transmission, representing an incremental improvement over existing deep compression models.
The paper tackled the problem of optimizing image compression networks for perceptual quality by introducing a proxy network, ProxIQA, to mimic perceptual models as a loss layer, resulting in up to a 31% bitrate reduction over MSE optimization at a specified VMAF quality level.
The use of $\ell_p$ $(p=1,2)$ norms has largely dominated the measurement of loss in neural networks due to their simplicity and analytical properties. However, when used to assess the loss of visual information, these simple norms are not very consistent with human perception. Here, we describe a different "proximal" approach to optimize image analysis networks against quantitative perceptual models. Specifically, we construct a proxy network, broadly termed ProxIQA, which mimics the perceptual model while serving as a loss layer of the network. We experimentally demonstrate how this optimization framework can be applied to train an end-to-end optimized image compression network. By building on top of an existing deep image compression model, we are able to demonstrate a bitrate reduction of as much as $31\%$ over MSE optimization, given a specified perceptual quality (VMAF) level.