Quantitative Assessment of Transformation Based Satellite Image Pan-sharpening Algorithms

Background and Objectives: Pan-sharpening algorithms integrate the spectral capabilities of the multispectral imagery with the spatial details of the panchromatic one to obtain a product with confident spectral and spatial resolutions. Due to the large diversities in the utilized pan-sharpening algorithms, occurring spatial and spectral deviations in their results should be recognized by performing the quantitative assessment analysis. Methods: In this research, the pan-sharpened images from PCA, IHS, and Gram-Schmidt transformation based algorithms are evaluated for the multi-spectral and panchromatic images fusion of Landsat-8 OLI sensor (medium scale resolution satellite) and WorldView-2 (high-resolution satellite). Quantitative analysis is performed on the pan-sharpened products based on the Per-Pixel Deviation (PPD) measure for spectral deviation analysis and high-pass filter and edge extraction measures for analyzing the spatial correlations. Moreover, entropy and standard deviation quantitative evaluation measures are also utilized based on the pan-sharpened image content. Results: Quantitative analysis represents that increasing the spatial resolution of the utilized remote sensing data has direct impacts on the spectral, spatial, and content-based characteristics of the generated Pan-sharpened products. Gram-Schmidt transformation based pan-sharpening method has the least spectral deviations in both WorldView-2 and Landsat-8 satellite images. But, the amount of spectral, spatial and content-based quantitative measures of PCA and IHS are changing with various spatial resolutions. Conclusion: it can be said that Gram-Schmidt pan-sharpening method has the best performance in both medium-scale and high-resolution data sets based on the spectral, spatial, and content quantitative evaluation results. The IHS pan-sharpening method has better performance than the PCA method in Landsat-8 OLI data. But, by increasing the spatial resolution of the data, PCA generates pan-sharpened products with better spectral, spatial, and content based quantitative evaluation results.