Abstract: In Earth observation applications, remote sensing image fusion has been proved to be an important means to enhance the spatial and spectral quality of the satellite image. Pansharpening is being developed as an effective method among different image fusion techniques aimed at merging the high spatial resolution panchromatic (PAN) image with the low spatial resolution multispectral (MS) data to produce high-resolution multispectral image. Nevertheless, it is still a huge challenge to obtain a balance between spatial enhancement and spectral preservation in conventional pansharpening methods. In this chapter, an extensive study of adaptive pansharpening techniques for high resolution remote sensing image fusion has been presented with special focus on Multiresolution Analysis (MRA) based adaptive pansharpening and Adaptive Hybrid Pansharpening (AHP) techniques. The proposed MRA-based method uses Gaussian-filter based spatial detail extraction, adaptive intensity estimation using the Levenberg–Marquardt (LM) learning algorithm, covariance-based gain factor computation, and adaptive detail injection to improve spatial resolution while maintaining the spectral information. In addition, an adaptive hybrid pansharpening method based on histogram matching, dual-level high-frequency detail extraction, adaptive fusion parameter estimation and local smoothing operations is proposed to further enhance the fusion effect. To assess this, satellite data representative of various geographical features (e.g., vegetation, water bodies, coastal, and urban areas) from the IKONOS and QuickBird satellites was used in experimental evaluation. The proposed techniques were compared with a few conventional and advanced pansharpening techniques, such as PCA, HPF, FIHS, AWLP, PRACS, conventional hybrid pansharpening and MTF-Wiener filtering. The quantitative results, measured using Relative Average Spectral Error (RASE), Erreur Relative Globale Adimensionnelle de Synthèse (ERGAS), Correlation Coefficient (CC), Average Quality Index (Q-average), and Average Gradient (AG), indicate that the proposed adaptive methods are capable of yielding better spectral preservation, spatial detail representation, and overall fusion quality. The adaptive hybrid pansharpening technique shows the most satisfactory results for both IKONOS and QuickBird data. The proposed adaptive fusion schemes provide efficient and effective image fusion solutions for high resolution remote sensing image fusion and have great potential applications in fields such as urban planning, environmental monitoring, precision agriculture, disaster management, and Earth observation system.

Page: 01-24