Adaptive ultrasonic speckle reduction based on the slope-facet model

The flat-facet model has been implicitly assumed for the structure of the image surface by most conventional speckle-reduction algorithms. However, this model is rarely found in a real ultrasound (US) image. To preserve the higher order structures and to capture the spatially variant property of the speckle, a new adaptive speckle-reduction algorithm, called the symmetrical speckle-reduction filter (SSRF), was developed based on the slope-facet model. The basic idea of the SSRF was to estimate the uncorrupted signal on the largest symmetrical slope facet centered at each target pixel. The symmetry constraint ensured the correctness of the mean value. An empirical speckle model was incorporated to account for the nature of the speckle in US image. A two-stage despeckling strategy was employed to enhance the statistical reliability of each estimate by forming a union of a set of symmetrical despeckling windows. The proposed SSRF algorithm was compared with two filtered-based and one wavelet-based approaches and the experimental results showed that the proposed SSRF outperformed these three previous approaches in both the synthetic images and the clinical US images tested in this study.