Visual Cryptography: A Comprehensive Survey of Schemes, Evaluation Metrics, and Emerging Research Directions

Visual cryptography (VC) constitutes a vibrant and quickly advancing sub-discipline of information security in which a confidential image is converted into either noise-resembling or semantically meaningful shares; the secret is subsequently retrieved either through the human visual apparatus or via elementary Boolean operations, without recourse to computational decryption. Although the topic has matured over roughly three decades, the corresponding literature is splintered across several sub-paradigms, and the most recent developments—propelled by deep neural networks, QR-code encoding, blockchain ledgers, and IoT ecosystems—have outstripped the coverage of existing survey articles. This paper assembles a consolidated review of VC that catalogues 77 distinct VC scheme and system references spanning 1995–2026, together with 10 foundational, image-quality-metric, and survey works dating back to 1979, and organizes them along six orthogonal axes: threshold, progressive, verifiable, multi-tone, multi-secret, and learning-based VC. Four principal contributions are made. First, a precise mathematical scaffolding is laid out, comprising the formal access-structure description, the construction of basis matrices, and the contrast–security inequalities that every well-formed VC scheme must satisfy. Second, a structured catalogue, assembled under an explicitly stated review methodology, is provided of both objective measures (NRM, BCR, BER, SSIM, F-measure, DRDM) and subjective indicators used throughout the field, together with the design trade-offs they impose on scheme construction. Third, an indicative cross-paradigm summary of author-reported performance figures (pixel expansion, contrast, PSNR, SSIM, and recovery modality) is compiled for nine representative schemes drawn from the literature; because these figures originate from heterogeneous test images, datasets, and metric conventions, the summary is presented as an orientation aid rather than as a controlled benchmark. We are not aware of a comparable cross-paradigm tabulation in any prior VC survey. Fourth, nine forward-looking research themes are identified: share generation through deep learning, quantum-secure VC, QR-driven and mobile VC, blockchain- and IoT-coupled VC, telemedicine, cloud and social-network privacy, adversarially-resilient cheater detection, standardized benchmarking, and metaheuristic optimization of VC parameters. The survey closes by synthesizing a set of open problems intended to direct research over the near term.