An Advanced Recomposition-Based Displaying Technique: Maximizing Image Reconstruction for Virtual Museum Applications

We present an advanced recomposition-based displaying technique designed to optimize intelligent scene retargeting in hybrid-reality environments, with two key contributions: Geometry-preserving and Human-inspired Active Detection (GHAD) and time-sensitive feature selection. GHAD progressively constructs Gaze Shift Paths (GSPs), aligning image processing with human gaze dynamics to maximize image reconstruction accuracy, while prioritizing key visual elements based on human attention patterns. The time-sensitive feature selection utilizes the BING objectness metric to identify and prioritize the most relevant features from multimodal data sources, ensuring efficient extraction and preserving exhibit content. These methods, combined with a multi-layer aggregation algorithm that encodes deep feature representations in a Gaussian Mixture Model (GMM), enable seamless scene reconstruction with improved precision. Empirical evaluations, including user studies, demonstrate the techniqueąŕs superiority, achieving 3.9% to 5.0% higher precision on six scenery sets and reducing testing time by 50%. The approach effectively balances algorithmic precision with human-centered aesthetics, advancing AI-driven scene analysis and visual recomposition, while enhancing interactivity and immersion for a more engaging and adaptive user experience.