Speckle filtering of ultrasonic images using a modified non local-based algorithm

The speckle reduces the contrast and obscures important details in ultrasound images, which limits the accuracy in clinical diagnose. Many methods have been developed to reduce the speckle. However, they suffer from several limitations, difficult to reach a balance between the noise attenuation and edge preserving. Recently, a newly developed filter, namely Non-Local Means (NL-means), works well in the image denoising, jointly enhancing the edge information. Since the NL-means filter is originally designed for Gaussian noise removal, it cannot be directly applied to ultrasonic image despeckling with the Rayleigh distribution noise. In this paper, we proposed a modified non local-based (MNL) filter to adapt for the speckle reduction. Experiments were carried out on synthetic images to find optimum parameters. Besides, this filter has been compared with other six state-of-the-art denoising methods on synthetic data and real ultrasonic images. Results show that the MNL method is capable of effectively reducing the speckle noise while well preserving tissue boundaries for ultrasonic images.