Search

In this paper, a novel multiscale speckle suppression method is proposed for ultrasonic images. An original image is transformed into a multiscale curvelet domain and the thresholding estimators are derived for processing the curvelet coefficients of detail subbands by successively combining the SURE-LET approach with the neighboring windows. In order to reduce the curvelet-like artifacts, the shrinkage results are further processed by a nonlinear diffusion scheme. We test the performance of this new technique using synthetic speckle images and real ultrasound images. The experimental results demonstrated, in comparison to other competitive existing methods, this technique could effectively eliminate the speckle noise, while preserving the detailed structures.