A fusion-based despeckling approach for multi-temporal SAR images using non-adaptive diffusion and thresholding method

Synthetic aperture radar images often suffer from speckle noise and blur, which can affect their reliability. This research proposes a new method that combines the speckle reducing anisotropic diffusion filter with the non-data adaptive transform and uses the Bayesian shrinkage rule for N-level decomposition. The effectiveness of this approach is analyzed by testing it on both synthetic and real SAR data sets, using visual and quality assessment metrics. The results indicate that the proposed approach significantly outperforms traditional filters. For instance, it achieves a PSNR of 51.8562 dB, surpassing log compression's 16.0555 dB for S-1 images. Additionally, it attains a UIQI of 0.98998 for S-2 images, exceeding the Lee filter's 0.68765 and an SSIM of 0.96897 for S-6 images, outperforming the Kuan filter's 0.48741 Furthermore, for S-8 images, the method's SNR of 47.0879 dB markedly exceeds Frost diffusion's 27.2356 dB. These findings demonstrate the method's superior capability in noise reduction.