Research On Despeckling For Ultrasound Images Based On Anisotropic Diffusion

Ultrasonography is an important medical imaging technique, which is widely used in clinical examination and diagnosis. The application of ultrasonography has been developed rapidly in the past few decades in that it produces relatively low-cost, noninvasive and easily accepted ultrasound images. However, ultrasound imaging process tends to introduce a form of multiplicative noise called speckle, which gives poor resolution and masks image details. The presence of speckle noise brings clinician a great distress for the analysis and diagnosis of ultrasound images.Speckle is not a certain and genuine noise. There is no consensus on a unified way to learn and interpret this undesired phenomenon. Although lots of filters have been proposed to suppressing speckle, it remains a delicate and hot issue in the field of ultrasound image processing. Therefore, this paper focuses on the study of coping with speckle noise for ultrasound images by means of anisotropic diffusion model.Firstly, this paper expounds the significance of reducing speckle noise by anisotropic diffusion. Secondly, the basic principle of ultrasound imaging is described, the cause and model of speckle noise through this paper are proposed and the quality assessment indices are introduced to compare the performance. Thirdly, several typical and representative anisotropic diffusion methods are presented and the corresponding performance is discussed respectively. In the end, this paper proposed two modified anisotropic diffusion methods as follows:1. Conventional anisotropic diffusion models use a template of four closest neighbors of the reference pixel under analysis for evaluating the instantaneous coefficient of variation, while this term can be calculated more accurately by applying a larger-sized template. Experimental results show more details and features related to the original image are preserved, which provide more meaningful diagnostic information for clinician. 2. When undergo anisotropic diffusion filtering with a sigmoid diffusive coefficient, the resulting images are good candidates for total variation denoising. Thus, this paper proposes a hybrid method for despeckling in ultrasonography. Experimental results indicate the proposed method reveals sharper boundaries from the visual effect than the other methods.