A SAR Image-Despeckling Method Based on HOSVD Using Tensor Patches

Coherent imaging systems, such as synthetic aperture radar (SAR), often suffer from granular speckle noise due to inherent defects, which can make interpretation challenging. Although numerous despeckling methods have been proposed in the past three decades, SAR image remains a challenging task. With extensive use of non-local self-similarity, under framework become increasingly mature. However, effectively utilizing patch similarities key problem despeckling. This paper proposes three-dimensional (3D) method based on searching for similar patches and applying high-order singular value decomposition (HOSVD) theory better utilize high-dimensional information patches. Specifically, extends two-dimensional (2D) 3D using tensor A new, patch-searching measure criterion is used classify patches, are stacked into tensors. Lastly, iterative adaptive weighted cyclic approximation HOSVD method. Experimental results demonstrate that not only reduces but also preserves fine details.