Synthetic models of ultrasound image formation for speckle noise simulation and analysis

摘要

Speckle noise is the primary cause of degradation of quality, resolution and contrast in ultrasound (US) images. Speckle in ultrasound B-mode images is caused by additive and destructive interference of ultrasound signals received from scatterers. Methods for analysing and reducing noise in US images require accurate models of image formation that can generate ground truth data. Such synthetic images that have the essential noise characteristics of real ultrasound images would be valuable for testing and evaluation of speckle reduction algorithms. This paper introduces three sampling models: radial polar, uniform grid and radial uniform that could be used for generating synthetic images. The paper also outlines the implementation aspects using pseudo-codes, and provides a comparative analysis between the proposed models. Experimental results showing variations in noise features with model parameters are also given.