Adaptive Bilateral Filtering for Despeckling of Medical Ultrasound Images

In this paper, an adaptive approach of bilateral filtering is introduced for the despeckling of medical ultrasound images. The range parameter is estimated from intensity homogeneity measurements. For each pixel, the measurements are carried out utilizing its local neighbors considering different directions. The range parameter is then estimated from the variance of the most homogeneous blocks and thus, automatically gets adapted according to the variations of the speckle noise. Experiments performed on synthetically-speckled images reveal that the proposed method outperforms several recently introduced despeckling techniques in terms of the signal-to-noise ratio and structural similarity index with a better preservation of image structures. It is shown that the proposed method improves the visual quality of ultrasound images by removing mostly noise while retaining the diagnostically important image details.