Enhanced Visual Retrieval and Decision Support

Abstract

The exponential growth of visual data across digital platforms, industrial monitoring, and healthcare applications has intensified the demand for intelligent and interpretable image retrieval systems. Traditional Content-Based Image Retrieval (CBIR) methods, based on handcrafted or convolutional descriptors, have laid the foundation for visual search but remain limited by low semantic interpretability, domain dependency, and restricted adaptability. In contrast, the rise of Generative Artificial Intelligence (AI) has introduced a new paradigm that redefines image retrieval as a process of semantic reasoning, synthesis, and decision support rather than simple feature matching. This paper presents a comprehensive analysis of recent advances in generative AI–driven retrieval architectures, emphasizing their ability to bridge the semantic gap, enhance multimodal understanding, and provide explainable results. The review demonstrates that generative approaches consistently improve retrieval accuracy, robustness to domain variations, and interpretability of results, outperforming traditional CBIR in multiple application scenarios. Beyond performance evaluation, this work contributes to the conceptual unification of generative retrieval systems and positions them as intelligent decision-support tools for applications in medical imaging, industrial inspection, smart surveillance, and event understanding. The findings highlight a decisive shift from perception-driven to reasoning-driven retrieval systems, underscoring the potential for adaptive, transparent, and explainable visual search frameworks that can support complex decision-making across domains.