A hybrid scaling factor for speckle reducing anisotropic diffusion

Abstract

Classical Speckle Reducing Anisotropic Diffusion (SRAD) is a noise filtering method tailored to speckle reduction in digital images. It is well known that SRAD has a tendency to produce dislocated and un-sharp edges. This property of SRAD is highly attributed to its reliance on a homogeneous image region, selected initially, for a scaling factor calculation. Moreover, this scaling factor calculation strategy requires the interaction of an experienced user. To reduce the reliance on an initially selected perfect/near-perfect homogeneous region, we propose a hybrid scaling factor that reduces edge dislocation and preserves the sharpness of edges. The proposed scaling factor selection strategy uses a ratio-based edge detection technique for an estimation of the homogeneity of the initially selected region.