Studying the spatial spectral fusion models for remote sensing images

Abstract Image fusion is the tactic of collecting two or more distinct imagery to produce a modern imagery using a model to learn more and good details about a subject. For many applications, the usage of freely available satellite imagery as Landsat 8 (L8) and Sentinel 2 (S2) is yet essential. In this study, the port said governorate was covered by the fusion of a 30 m spatial resolution L8 level- 2 and a 10 m spatial resolution S2 level 2 A and the ismailia city was covered by the fusion of a 43 cm spatial resolution high resolution (HR) and a 10 m spatial resolution S2 Level 2 A. Applying the Gram-Schmidt (GS), nearest neighbor diffuse, brovey, intensity-hue-saturation, and simple mean algorithms. The main aim of this paper to improve the spatial resolution of L8 (by pan sharpening with S2) and the spatial resolution of S2 (by pan sharpening with HR). The fused images are assessed using high-quality image techniques as error relative global average squared, root mean squared error, entropy, structural similarity index measure, and correlation coefficient. The outcomes demonstrated that the GS method based on the red band of S2 (band 4) has the preferable results for fusion between L8 and S2 for port said governorate and brovey method has the preferable results for fusion between HR and S2 for ismailia city. Following these results, the study's following phase examined how various scale (S) parameters affected the image segmentation process. Segmentation is an essential step in the conversion of pixel-depended image analysis to object-depended image analysis. The outcomes demonstrate that the preferable values for the GS fusion method, depend on the S2 red band (band 4), are about S factor 70 for fusion between L8 and S2 and about S factor 50 and 60 for fusion between HR and S2.