MemOVCD: Training-Free Open-Vocabulary Change Detection via Cross-Temporal Memory Reasoning and Global-Local Adaptive Rectification

2026-04-29 · Source: Computer Vision and Pattern Recognition · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Computer Vision · Depth: Expert, quick

Summary

MemOVCD is a training-free open-vocabulary change detection framework designed for bi-temporal remote sensing images, addressing the challenge of identifying semantic changes without predefined categories. Existing methods often process timestamps independently or interact only at the final comparison, leading to insufficient temporal coupling and fragmented change regions in high-resolution images. MemOVCD tackles these limitations by reformulating change detection as a two-frame tracking problem, employing weighted bidirectional propagation to aggregate semantic evidence across time. It also utilizes histogram-aligned transition frames to smooth abrupt appearance changes and stabilize memory propagation. Furthermore, a global-local adaptive rectification strategy fuses local and global-view predictions to enhance spatial consistency while retaining fine-grained details. Experiments on five benchmarks confirm MemOVCD's effectiveness and generalization across diverse open-vocabulary settings for two change detection tasks.

Key takeaway

For research scientists developing open-vocabulary change detection systems, MemOVCD's approach offers a robust, training-free alternative. You should consider integrating cross-temporal memory reasoning and global-local adaptive rectification into your models to improve semantic change identification and spatial consistency, especially when dealing with high-resolution remote sensing imagery and diverse open-vocabulary scenarios.

Key insights

MemOVCD improves open-vocabulary change detection by integrating cross-temporal memory reasoning and global-local adaptive rectification.

Principles

• Reformulate change detection as tracking.
• Aggregate semantic evidence bidirectionally.
• Smooth abrupt changes with transition frames.

Method

MemOVCD uses weighted bidirectional propagation for cross-temporal memory reasoning, constructs histogram-aligned transition frames, and applies a global-local adaptive rectification strategy to fuse predictions for improved spatial consistency.

In practice

• Apply two-frame tracking to bi-temporal images.
• Use histogram alignment for temporal smoothing.
• Combine global and local predictions.

Topics

• Open-Vocabulary Change Detection
• MemOVCD
• Cross-Temporal Memory Reasoning
• Global-Local Adaptive Rectification
• Remote Sensing Images

Best for: Research Scientist, AI Scientist, Computer Vision Engineer

Related on AIssential

• ReA-OVCD: Reliability-Aware Open-Vocabulary Change Detection via Semantic and Spatial Refinement — ReA-OVCD refines open-vocabulary change detection by collaboratively addressing semantic ambiguity and spatial inconsistency through a training-free, two-module approach.
• Modeling visual memorability assessment with autoencoders reveals characteristics of memorable images — Autoencoder-based models can identify visual features that predict image memorability.
• RSRCC: A Remote Sensing Regional Change Comprehension Benchmark Constructed via Retrieval-Augmented Best-of-N Ranking — RSRCC is a new benchmark for fine-grained, localized remote sensing change question-answering.
• SAM 3: Concept-Based Visual Understanding and Segmentation — SAM 3 unifies detection, segmentation, and tracking of open-vocabulary concepts using text or visual prompts.
• An Open-Source Benchmark and Baseline for Multi-temporal Referring Segmentation — Multi-temporal Referring Segmentation (MTRS) is a new task and benchmark for language-guided temporal change detection in images.
• Take a Peek: Efficient Encoder Adaptation for Few-Shot Semantic Segmentation via LoRA — Take a Peek (TaP) efficiently adapts FSS encoders to novel classes using Low-Rank Adaptation (LoRA) on support sets.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Computer Vision and Pattern Recognition.