specklenoise

Specklenoise is a term used to describe the granular, random pattern that appears in images produced by coherent imaging systems, caused by interference of coherent waves scattered from rough or inhomogeneous media. It is often treated as a multiplicative noise that degrades image quality by reducing contrast and effective resolution.

In many imaging contexts, specklenoise can be modeled as a multiplicative component: I = J · S, where I is the observed intensity, J is the underlying reflectivity or scene brightness, and S represents the speckle. In the log domain, this becomes log I = log J + log S, yielding an additive noise term with characteristic statistics. The statistical form of speckle depends on the modality: fully developed speckle in synthetic aperture radar (SAR) and ultrasound is often described by gamma or Rayleigh-like distributions for intensity, with spatial correlations determined by the imaging system and the scatterer field.

Specklenoise is common in ultrasound imaging, SAR, laser speckle in optical coherence imaging, and other coherent modalities. It can obscure fine details, reduce signal-to-noise and contrast-to-noise ratios, and complicate quantitative analysis. Characteristics such as decorrelation with time, angle, or wavelength can influence speckle patterns, informing mitigation strategies.

Despeckling approaches aim to suppress specklenoise while preserving structural details. Methods include spatial and temporal filtering (eg, Lee, Frost), nonlocal or multiscale techniques, wavelet-based and BM3D-inspired algorithms, and modern deep learning models. Acquisition strategies such as angular or polarization diversity, higher sampling, and averaging can also mitigate specklenoise.