This paper presents a new texture based anisotropic diffusion method for real ultrasound image despeckling. Texture information is obtained by a real ultrasound image model. Unlike traditional anisotropic diffusion methods usually taking image gradient as a diffusion index, we take the image texture as a new diffusion index. The results comparing our new method with others on both simulated image and real ultrasound images are reported, and our method shows the superiority in keeping important features of real ultrasound images.

Speckle noise is widely existence in coherent imaging systems, such as synthetic aperture radar, sonar, ultrasound and laser imaging, and is commonly described as signal correlated. In this paper, we focus on speckle reduction problem in real ultrasound image. Unlike traditional anisotropic diffusion methods usually taking image gradient as a diffusion index, in this paper, we present a new texture based anisotropic diffusion method for speckle reduction in real ultrasound image. The results comparing our new method with other well known methods on both synthetic images and real ultrasound images are reported to show the superiority of our method in keeping important features of real ultrasound images.