Improving Statistical Fidelity for Neural Image Compression with Implicit Local Likelihood Models
This work addresses the issue of statistical discrepancies in compressed images for applications requiring high visual quality, representing an incremental improvement over existing adversarial methods.
The paper tackles the problem of statistical fidelity loss in neural image compression at low bitrates, which causes blurring, by introducing a non-binary discriminator conditioned on quantized local representations, achieving the same FID as the state-of-the-art HiFiC model with 30-40% fewer bits on the CLIC2020 dataset.
Lossy image compression aims to represent images in as few bits as possible while maintaining fidelity to the original. Theoretical results indicate that optimizing distortion metrics such as PSNR or MS-SSIM necessarily leads to a discrepancy in the statistics of original images from those of reconstructions, in particular at low bitrates, often manifested by the blurring of the compressed images. Previous work has leveraged adversarial discriminators to improve statistical fidelity. Yet these binary discriminators adopted from generative modeling tasks may not be ideal for image compression. In this paper, we introduce a non-binary discriminator that is conditioned on quantized local image representations obtained via VQ-VAE autoencoders. Our evaluations on the CLIC2020, DIV2K and Kodak datasets show that our discriminator is more effective for jointly optimizing distortion (e.g., PSNR) and statistical fidelity (e.g., FID) than the PatchGAN of the state-of-the-art HiFiC model. On CLIC2020, we obtain the same FID as HiFiC with 30-40\% fewer bits.