Training-Free Rate-Distortion-Perception Traversal With Diffusion
This work addresses the problem of adaptive, perception-aware compression for applications requiring flexible and efficient compression methods, providing an incremental yet significant improvement over existing neural compression methods.
The authors tackled the problem of navigating the rate-distortion-perception tradeoff in lossy compression, achieving a flexible and effective framework that can traverse the entire tradeoff surface without requiring retraining, with empirical results demonstrating its effectiveness across multiple datasets. The framework achieves optimal distortion-perception tradeoff under AWGN observations and optimal RDP function in the Gaussian case.
The rate-distortion-perception (RDP) tradeoff characterizes the fundamental limits of lossy compression by jointly considering bitrate, reconstruction fidelity, and perceptual quality. While recent neural compression methods have improved perceptual performance, they typically operate at fixed points on the RDP surface, requiring retraining to target different tradeoffs. In this work, we propose a training-free framework that leverages pre-trained diffusion models to traverse the entire RDP surface. Our approach integrates a reverse channel coding (RCC) module with a novel score-scaled probability flow ODE decoder. We theoretically prove that the proposed diffusion decoder is optimal for the distortion-perception tradeoff under AWGN observations and that the overall framework with the RCC module achieves the optimal RDP function in the Gaussian case. Empirical results across multiple datasets demonstrate the framework's flexibility and effectiveness in navigating the ternary RDP tradeoff using pre-trained diffusion models. Our results establish a practical and theoretically grounded approach to adaptive, perception-aware compression.