ME-IQA: Memory-Enhanced Image Quality Assessment via Re-Ranking
For practitioners of IQA, this work mitigates a key limitation of reasoning-based VLMs (discrete collapse) with a plug-and-play test-time method, but the gains are incremental as it builds on existing VLM frameworks.
ME-IQA addresses the discrete collapse problem in reasoning-based image quality assessment (IQA) where scalar scores from VLMs lack sensitivity. By introducing a memory-enhanced re-ranking framework that retrieves neighbors, computes pairwise preferences, and fuses evidence, it achieves denser, distortion-sensitive predictions and consistent gains over baselines across multiple IQA benchmarks.
Reasoning-induced vision-language models (VLMs) advance image quality assessment (IQA) with textual reasoning, yet their scalar scores often lack sensitivity and collapse to a few values, so-called discrete collapse. We introduce ME-IQA, a plug-and-play, test-time memory-enhanced re-ranking framework. It (i) builds a memory bank and retrieves semantically and perceptually aligned neighbors using reasoning summaries, (ii) reframes the VLM as a probabilistic comparator to obtain pairwise preference probabilities and fuse this ordinal evidence with the initial score under Thurstone's Case V model, and (iii) performs gated reflection and consolidates memory to improve future decisions. This yields denser, distortion-sensitive predictions and mitigates discrete collapse. Experiments across multiple IQA benchmarks show consistent gains over strong reasoning-induced VLM baselines, existing non-reasoning IQA methods, and test-time scaling alternatives.