Abstract
Speckle noise is a fundamental challenge in coherent imaging systems,significantly degrading image quality. Over the past decades, numerousdespeckling algorithms have been developed for applications such as SyntheticAperture Radar (SAR) and digital holography. In this paper, we aim to establisha theoretically grounded approach to despeckling. We propose a methodapplicable to general structured stationary stochastic sources. We demonstratethe effectiveness of the proposed method on piecewise constant sources.Additionally, we theoretically derive a lower bound on the despecklingperformance for such sources. The proposed depseckler applied to the 1-Markovstructured sources achieves better reconstruction performance with no strongsimplification of the ground truth signal model or speckle noise.