Good, Cheap, and Fast: Overfitted Image Compression with Wasserstein Distortion

2024-11-30

arXiv.org

AI Summary

To address the challenge of simultaneously achieving high reconstruction quality and low computational complexity in perceptual image compression, this paper proposes C3, a lightweight overfitting codec that abandons generative modeling and instead adopts Wasserstein distortion (WD) as the end-to-end optimization objective—the first such use in learned image compression. C3 reduces decoding computational cost to less than 1% of mainstream generative models (measured in MACs) while matching HiFiC’s perceptual quality. A key contribution is the empirical validation of WD as the strongest known proxy for human subjective quality: it achieves a Pearson correlation of 94.2% with Elo ratings—substantially surpassing LPIPS, DISTS, and MS-SSIM. Experiments follow the standard Elo evaluation protocol, demonstrating that WD-driven optimization delivers superior perceptual fidelity and deployment efficiency.

Technology Category

• Computer Vision: Learning & Optimization for CV
• Machine Learning: Learning on the Edge & Model Compression
• Cognitive Modeling & Cognitive Systems: Computational Creativity

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Systems and Infrastructure for Web, Mobile and WoT: Web performance, measurement, and characterization

Abstract

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.

Problem

• Balancing visual quality and computational efficiency in image compression
• Optimizing overfitted image codec with Wasserstein Distortion for better perception
• Evaluating perceptual metrics for human-rated image reconstruction quality

Innovation

• Optimizing overfitted codec for Wasserstein Distortion
• Focusing on visual perception not data distribution
• Achieving high quality with low computational cost