Speckle reduction of ultrasound images based on locally regularized Savitzky-Golay filters

This paper describes a new filtering algorithm developed for real-time speckle noise reduction of medical ultrasound images. The new filter, referred to as the locally regularized Savitzky-Golay (LRSG) filter, is the two-dimensional Savitzky-Golay (2-D SG) filter, enhanced with a mechanism for adapting the degree of smoothing to match both the strength of speckle and image structures in each local regions of the image. The results comparing the new filter algorithm with the Adaptive Speckle Reduction and Adaptive Weighted Median filters on a ultrasound phantom and a real ultrasound thyroid image are reported.