Nonlinear diffusions and structure tensor in modelling to speckle noise reduction and edge enhancement in ultrasound images.

Nguyễn Hải Hà, Phạm Trần Nhu


Speckle noise generally affects medical ultrasound images quality, and tends to reduce the image resolution and contrast. Speckle noise is an inherent property in which the images are formed under random interference between the coherent natures returns of a transmitted waveform that cause from scattering phenomenon. Many solutions have been proposed so far to remove speckle noise without loosing the edge information in images. This paper proposes two diffusion processes, in which isotropic diffusion solves the problem of speckle noise reduction and anisotropic diffusion enhances edges and local details in ultrasound images. Both the cases are controlled by the nonlinear diffusion and structure tensor model. The proposed model combines between the regularized nonlinear diffusion process with the mean curvature motion equation and structure tensor. The model performs simultaneous speckle noise reduction in homogeneous region, structure enhancement in inhomogeneous region using nonlinear diffusion based on local variations of the gradient orientation of an image. The paper also presents experimental results carried out on ultrasound images affected by speckle noise for illustrating the effectiveness of the proposed model.


nonlinear diffusions, structure tensors, speckle noise, image edge enhancement, ultrasound image.


Journal of Computer Science and Cybernetics ISSN: 1813-9663

Published by Vietnam Academy of Science and Technology