Panoptic Pairwise Distortion Graph

TL;DR

This paper introduces a novel structured approach to compare image pairs by representing regions and distortions as interpretable graphs, along with datasets, benchmarks, and an architecture for this task.

Contribution

It presents a new task of Distortion Graph, a region-level dataset PandaSet, a benchmark PandaBench, and an architecture Panda for structured pairwise image assessment.

Findings

State-of-the-art multimodal models struggle with region-level degradation understanding.

Training on PandaSet or using Distortion Graph prompts improves region-wise distortion comprehension.

The proposed approach enables fine-grained, structured comparison of image pairs.

Abstract

In this work, we introduce a new perspective on comparative image assessment by representing an image pair as a structured composition of its regions. In contrast, existing methods focus on whole image analysis, while implicitly relying on region-level understanding. We extend the intra-image notion of a scene graph to inter-image, and propose a novel task of Distortion Graph (DG). DG treats paired images as a structured topology grounded in regions, and represents dense degradation information such as distortion type, severity, comparison and quality score in a compact interpretable graph structure. To realize the task of learning a distortion graph, we contribute (i) a region-level dataset, PandaSet, (ii) a benchmark suite, PandaBench, with varying region-level difficulty, and (iii) an efficient architecture, Panda, to generate distortion graphs. We demonstrate that PandaBench poses a significant challenge for state-of-the-art multimodal large language models (MLLMs) as they fail to understand region-level degradations even when fed with explicit region cues. We show that training on PandaSet or prompting with DG elicits region-wise distortion understanding, opening a new direction for fine-grained, structured pairwise image assessment.