A Hybrid Approach for Speckle Reduction in Ultrasound

Abstract

Speckle noise content present in ultrasound (US) images degrades the image contrast and makes image interpretation difficult. In this paper, a hybrid speckle reduction method has been proposed in which wavelet transform, 2D Wiener filter, and morphological operations are combined. The proposed method reduces speckle noise efficiently and enhances the US image. A comparison of the proposed method is made by utilizing classical speckle reduction filters including combinations of Fourier and homomorphic filters. For performance evaluation, we have used mean square error (MSE) along with peak signal-to-noise ratio (PSNR) for determining image quality, and signal-to-noise ratio (SNR) and normalized absolute error (NAE) techniques have been used for quantitative evaluation. Also, structural similarity index metric (SSIM) is used for qualitative evaluation of the US image. We have used synthetic and real US images in our proposed method for evaluation of performance.