Gradient Controlled Adaptive Non-Local Means Method for Speckle Reduction in Ultrasound Images

Speckle reduction is extremely important in the field of medical ultrasound (US) imaging. However, the extant despeckling methods of US images cannot provide satisfactory detail preservation or noise suppression, especially at high speckle corruption. To address this issue, a gradient controlled adaptive nonlocal means (GNLM) method has been put forward to remove speckle noise in US images. The GNLM method determines the similarity by utilizing the image pre-filtered by the optimized Bayesian NLM (OBNLM) method and adopting the pixel-wise adaptive decay parameters which are computed from the gradient of the preprocessed image. Extensive experiments show that the GNLM method can provide higher peak signal-to-noise ratio (PSNR) and structural similarity (SSIM) for the synthetic image as well as higher contrast-to-noise ratio (CNR) and equivalent number of looks (ENL) for four medical US images than several state-of-the-art despeckling methods. Subjective comparisons indicate that the GNLM method can suppress speckle noise, preserve small structures and enhance image contrast more effectively than the compared denoising methods.