Improved Chaotic Particle Swarm Optimization with Variance Based Novel Objective Function for Pansharpening in Remote Sensing

This paper proposes an improved pansharpening method using a Chaotic Particle Swarm Optimization (CPSO) framework, incorporating a novel variance-based objective function and fusion rule. To address limitations of conventional PSO, such as poor convergence and random initialization, we employ a chaotic logistic map for initializing particle positions, enhancing exploration and diversity. A modified ERGAS (Erreur Relative Globale Adimensionnelle de Synthèse) index, based on local variance between multispectral and fused images, is introduced to improve spatial detail preservation. Additionally, a variance-based Max fusion rule is applied to RGB and NIR bands, enhancing edge and spectral fidelity. Experimental evaluations on seven datasets from QuickBird and Landsat-7 ETM + demonstrate superior performance of the proposed method over existing techniques. Results show significant improvements in image quality metrics—ERGAS, SAM (Spectral Angle Mapper), RASE (Relative Average Spectral Error), RMSE (Root Mean Square Error), UIQI (Universal Image Quality Indexes), and CC (Correlation Coefficient)—highlighting the effectiveness of the method for remote sensing applications.