Successive refinement coding improves perception with common randomness
Successive Refinement Under Strong-Sense Perfect Perception
Information Theory
Summary
This paper studies how to compress information so that two users can recover it at different quality levels, focusing on making the recovered data look perfectly natural to human perception. The authors examine how adding a strong perception constraint affects compression, using a scenario where shared randomness is unlimited. They prove that for a simple type of data called a Bernoulli source, good compression is possible without harming perceptual quality. This work helps clarify how to balance data compression, quality, and natural appearance together.
What this means in practice
- •For image compression developers: Create multi-level image compressors that maintain high perceptual quality at each decoding stage using shared randomness for better progressive transmission.
- •For video streaming teams: Design streaming protocols where different clients receive video at varied qualities that all appear natural, enhancing user experience under bandwidth limits.
Authors
Yu Yang, Changhong Liu, Weijie Yuan, Lin Zhou
Abstract
We revisit a multiterminal lossy source coding problem named successive refinement and derive the rate-distortion-perception region under the strong-sense perfect perception constraint in the presence of unlimited common randomness. Specifically, in successive refinement, one aims to compress a source sequence and allows two distinct decoders to recover the source sequence at different distortion levels. By imposing the strong-sense perfect perception constraint, our results refine the previous result by analyzing the impact of the perceptual quality. Our achievability proof is inspired by output constrained lossy source coding and our converse proof adapts the proof steps of the standard successive refinement problem. Furthermore, we provide a numerical example of the Bernoulli source to illustrate our result and show that the Bernoulli source under Hamming distortion is successively refinable even with the strong-sense perfect perception constraint.