Pan-sharpening Through Weighted Total Generalized Variation Driven Spatial Prior and Shearlet Transform Regularization
摘要
The pivotal remote sensing applications like change detection, land use classicfication etc. desires the imaging products with robust geometric and color information components. The technological constraints limit the sensors in yielding the single image product with requisite spatial and spectral details. Pansharpening is a remote sensing image fusion mechanism that merge the structural details from Panchromatic image and spectral data from the multispectral image into a single product namely, high resolution multispectral (HRMS) image. Regardless of the availability of numerous fusion models, there is a demand for a comprehensive fusion pardagim that generates an imaging product with equitable spatial and spectral contents. In view of this, a pansharpening technique is proposed to intensify the resolution characteristics of the fused product. The proposed method relies on variational framework by employig the weighted total generalized variation and shearlet transform as priors fro spatial and spectral information contributions respectively. Finally, the Pansharpening is modeled as optimization problem and an efficient solver is used to realize the requisite HRMS image. Three representative datasets and five quality metrics are utilized to valuate the proposed method at original resolution and reduced resolution. The results certified that the proposed method yields promising results.