Good, Cheap, and Fast: Overfitted Image Compression with Wasserstein Distortion
This work addresses the computational inefficiency of learned image compression for practical applications, offering a more efficient alternative to generative models.
The paper tackles the high computational cost of generative image compression models by focusing on modeling visual perception rather than the data distribution, achieving a trade-off in visual quality and bit rate similar to HiFiC with less than 1% of the multiply-accumulate operations for decompression. It shows that optimizing an overfitted image codec for Wasserstein Distortion outperforms LPIPS as an objective and achieves over 94% Pearson correlation with human ratings.
Inspired by the success of generative image models, recent work on learned image compression increasingly focuses on better probabilistic models of the natural image distribution, leading to excellent image quality. This, however, comes at the expense of a computational complexity that is several orders of magnitude higher than today's commercial codecs, and thus prohibitive for most practical applications. With this paper, we demonstrate that by focusing on modeling visual perception rather than the data distribution, we can achieve a very good trade-off between visual quality and bit rate similar to "generative" compression models such as HiFiC, while requiring less than 1% of the multiply-accumulate operations (MACs) for decompression. We do this by optimizing C3, an overfitted image codec, for Wasserstein Distortion (WD), and evaluating the image reconstructions with a human rater study, showing that WD clearly outperforms LPIPS as an optimization objective. The study also reveals that WD outperforms other perceptual metrics such as LPIPS, DISTS, and MS-SSIM as a predictor of human ratings, remarkably achieving over 94% Pearson correlation with Elo scores.