Abstract Background: As the metaverse rapidly evolves into a highly immersive digital environment, the boundaries separating the physical and virtual worlds are becoming increasingly blurred. A foundational driving force behind this transformation is computer vision—a subset of artificial intelligence that empowers computers to process, analyze, and comprehend visual data in order to emulate human sight and interpretation. Objectives: This paper provides a comprehensive overview of current computer vision techniques and variations, exploring their integration into the metaverse and highlighting their critical importance across multiple industries. Methodology: The study reviews core deep learning and machine learning visual algorithms, categorizing basic computer vision techniques into image classification, object detection (via bounding boxes), and motion tracking. It evaluates the chronological four-step automated sequence through which machines recognize images: acquisition, interpretation, feature extraction, and pattern recognition. Findings: Computer vision is shown to be crucial for generating realistic 3D user environments and seamless human-virtual integration under challenging environmental variables. Key enabling functions include gesture/avatar recognition, spatial awareness, scene comprehension, content moderation, realistic non-player characters (NPCs), deep reinforcement learning (DRL) agents, and real-time pose estimation. The paper highlights transformative metaverse applications leveraging computer vision across diverse sectors, including gaming, social media, virtual tourism, inclusive education, medical training simulations, and e-commerce. Conclusion and Challenges: Despite its exponential growth, the deployment of computer vision in virtual environments faces notable technical hurdles, such as lag that disrupts user immersion. Additionally, severe ethical challenges persist, including deepfakes, algorithmic biases inherited from training data, and heightened privacy concerns surrounding surveillance and facial recognition. The paper concludes that developers and policymakers must prioritize responsible, non-discriminatory frameworks to balance technological innovation with ethical safeguards.
Mohamed et al. (2026) studied this question.