Denoise Auto-Encoder Based Speckle Reduction for RISAT-1 SAR Imagery

Abstract

Synthetic aperture radar (SAR) images can pass through cloud cover, dry particles, and haze except for heavy rainfall. Therefore, they are available in all climates, all the time. However, the SAR images are corrupted by speckle noise generated by coherent processing of SAR signal. In this paper, the denoise auto-encoder model is proposed to reduce the speckle noise in SAR images, and the performance of the auto-encoder model is compared with different spatial-domain adaptive filters viz. Lee, Frost, Enhanced Lee, Enhanced Frost. The performance of the proposed denoising encoder is assessed using visual analysis, and quantitative evaluation using metrics, viz. equivalent number of looks (ENL), speckle suppression index (SSI) and speckle suppression and mean preservation index (SMPI). The evaluation of the denoise auto-encoder reveals that its performance is better than spatial- domain adaptive filters.