Particle swarm optimization for multi-exposure image fusion incorporating dual guidance from visual quality and brightness detail metrics

Abstract

To overcome the limitations associated with fixed weight allocation in multi-exposure image fusion techniques— specifically the loss of details in both bright and dark regions and local overexposure—this study introduces a novel multi-exposure image fusion approach based on particle swarm optimization (PSO), guided by criteria emphasizing visual quality and enhancement of bright-dark details. The proposed method employs an image pyramid decomposition of multi-exposure inputs to extract Laplacian detail components across multiple scales, subsequently constructing a multidimensional feature weight model. Leveraging an objective function that integrates visual quality metrics with bright-dark detail enhancement, the PSO algorithm dynamically optimizes the weighting coefficients, enabling differentiated enhancement of details in luminance-extreme regions. The final fused image is produced through multi-scale reconstruction, yielding high-quality results. Experimental evaluations indicate that the proposed algorithm surpasses current state-of-the-art methods by delivering superior visual quality and better preservation of detail in both bright and dark areas. Furthermore, when applied to industrial inspection of mobile phone power adapters, the algorithm effectively accentuates defect features while suppressing reflective artifacts, demonstrating its potential utility in industrial visual inspection tasks involving high-reflectivity surfaces.