Purpose
Hyperspectral data are the most widely used remote sensing datasets. Hyperspectral Pan-Sharpening suffers from spectral distortion; the purpose of hyperspectral image fusion is to effectively inject the missing spatial detail into the HS image, while preserving the spectral information. Edge-preserving smoothing filters such as Guided Filter retains image edge and structure details while minimizing noise, gradient reversal, undershoot and overshoot artefacts. However, it exhibits halo artefacts.
Method
This paper introduces an innovative algorithm for panchromatic and hyperspectral image fusion. By employing the Adaptive Guided Filter, we enhance image sharpness and mitigate halo artefacts – objectionable counter shading around edges. This preserves image structure and aesthetic quality, maintaining the speed of Guided Filter.
Results
We have applied our technique on three Hyperspectral datasets such as DCMall, Salinas, and Moffett. Our technique has shown visually improved results for the halo artefacts over Guided Filter. We have established a comparison of our technique with Guided Filter in terms of fusion quality metrics such as cross correlation (CC), spectral angle mapper (SAM), root mean squared error (RMSE) and Erreur relative globale adimensionnelle de synthèse (ERGAS).
Conclusion
At smoothness (𝛔=2), our technique has shown better results for CC, SAM, RMSE, and ERGAS, indicating Adaptive Guided Filter perform better over Guided Filter in terms of retaining spatial quality and spectral quality.