Looks Too Good To Be True: An Information-Theoretic Analysis of Hallucinations in Generative Restoration Models

TL;DR

This paper uses information theory to analyze the fundamental tradeoff between perceptual quality and uncertainty in generative image restoration models, revealing inherent limitations in achieving both high perceptual quality and reliability.

Contribution

It introduces a theoretical framework linking perception, uncertainty, and distortion, proving that perfect perceptual quality necessarily involves increased uncertainty in generative models.

Findings

Inherent uncertainty in restoration grows with perceptual quality.

Achieving perfect perceptual quality requires at least twice the inherent uncertainty.

The perception-distortion tradeoff is fundamentally linked to an uncertainty-perception tradeoff.

Abstract

The pursuit of high perceptual quality in image restoration has driven the development of revolutionary generative models, capable of producing results often visually indistinguishable from real data. However, as their perceptual quality continues to improve, these models also exhibit a growing tendency to generate hallucinations - realistic-looking details that do not exist in the ground truth images. Hallucinations in these models create uncertainty about their reliability, raising major concerns about their practical application. This paper investigates this phenomenon through the lens of information theory, revealing a fundamental tradeoff between uncertainty and perception. We rigorously analyze the relationship between these two factors, proving that the global minimal uncertainty in generative models grows in tandem with perception. In particular, we define the inherent uncertainty of the restoration problem and show that attaining perfect perceptual quality entails at least twice this uncertainty. Additionally, we establish a relation between distortion, uncertainty and perception, through which we prove the aforementioned uncertainly-perception tradeoff induces the well-known perception-distortion tradeoff. We demonstrate our theoretical findings through experiments with super-resolution and inpainting algorithms. This work uncovers fundamental limitations of generative models in achieving both high perceptual quality and reliable predictions for image restoration. Thus, we aim to raise awareness among practitioners about this inherent tradeoff, empowering them to make informed decisions and potentially prioritize safety over perceptual performance.