Cross-Resolution Semantic Transfer for Robust Text-to-Image Retrieval in Low-Resolution Surveillance

2026-06-29

AI Summary

This work addresses the challenges of text-to-image person re-identification in real-world surveillance scenarios, where resolution discrepancies—particularly low resolution—lead to unreliable visual evidence and unstable ranking performance. To mitigate these issues, the authors propose CRST, a cross-resolution semantic transfer framework built upon the CLIP architecture. CRST introduces resolution-conditioned attention, text-guided feature refinement, and a cross-resolution rank distribution alignment (CR-RDA) mechanism, offering the first systematic approach to modeling and alleviating both evidence reliability collapse and rank distribution shift. Extensive experiments demonstrate that CRST improves Rank-1 accuracy and mAP by average margins of 5.7% and 5.3%, respectively, under ultra-low-resolution conditions on CUHK-PEDES, ICFG-PEDES, and RSTPReid benchmarks, while significantly stabilizing mixed-resolution retrieval performance without compromising high-resolution accuracy.

Abstract

Text-to-image person re-identification (TIPR) retrieves target persons using natural language descriptions. However, existing methods largely overlook resolution variance in real-world surveillance. They characterize cross-resolution TIPR through two coupled failure modes: Evidence Reliability Collapse (ERC), where degraded visual tokens become unreliable for grounding fine-grained text, and Ranking Distribution Drift (RDD), where mixed-resolution galleries distort similarity neighborhoods and destabilize retrieval rankings. To address this challenge, we propose Cross-Resolution Semantic Transfer (CRST), a CLIP-style framework with three modules: resolution-conditioned reasoning, text-guided refinement and CR-RDA. Resolution-conditioned reasoning estimates token reliability to suppress corrupted evidence. Text-guided refinement injects semantic priors to recover discriminative cues. CR-RDA transfers HR neighborhood geometry to stabilize LR ranking under mixed resolutions. Experiments on CUHK-PEDES, ICFG-PEDES, and RSTPReid show that CRST improves ultra-low-resolution Rank-1 and mAP on average by 5.7% and 5.3%, while stabilizing mixed-resolution retrieval without sacrificing high-resolution accuracy.The code will be made publicly available.

Problem

• text-to-image person re-identification
• low-resolution surveillance
• cross-resolution retrieval
• evidence reliability collapse
• ranking distribution drift

Innovation

• Cross-Resolution Semantic Transfer
• Text-to-Image Person Re-identification
• Resolution-Conditioned Reasoning
• Ranking Distribution Alignment
• Low-Resolution Surveillance