A nonlinear algorithm for speckle reduction

A nonlinear image processing algorithm called geometric filtering is demonstrated. This technique, originally developed for reducing the speckle in synthetic aperture radar imagery, is an iterative-rule based method which uses a pixel´s nearest neighbors to decide if an intensity modification is required. By applying the rule, rapid fluctuations in pixel intensity caused by speckle are more quickly eliminated than true image detail. The method has been applied directly to the video images of several commercial ultrasound imagers the preservation of lateral and axial resolution. The effect of the filtering method upon the statistical distribution of pixel intensities is demonstrated. The performance of the algorithm is quantified using histograms and contrast-speckle analysis on images obtained from simulated lesions in three contrast resolution phantoms. A series of static images, obtained using both linear and sector scanning, is presented