Application Of Coherence-Enhancing Diffusion Based On PDE In Medical US Image Processing

Ultrasound (US) has major advantages over many other imaging technologies because it is non-destructive, low cost and convenient, moreover it provides a real-time imaging, therefore it is widely used in medical diagnose field. However, because of its coherent property, US also has some disadvantages. It often has many strong speckle noises, low resolution, and fuzzy contour, which decreases US image's quality, especially, it covers up some special information in US images. Accordingly, these disadvantages not only affect accuracy of clinical diagnosis, but also lead to many difficulties on image post-processing of edge detection and feature extraction. Consequently, the processing of de-noising and enhancing on medical US image becomes a popular research subject in image processing field.Image de-noising and enhancing based on partial differential equation (PDE) is an important branch in image processing field, one of which is coherence-enhancing diffusion. Because of flexibility and characteristics of image structure processing, coherence-enhancing diffusion causes extensive concern. With gradually deepening of researching and continuous improvement of structure tensor, coherence-enhancing diffusion is more and more applied to medical US image processing. A large number of experiments show that when coherence-enhancing diffusion is used in medical US image processing, it can preserve the important structure information in the original US image, which is de-noised and enhanced at the same time. Furthermore, it can improve the visualization of the images. Obviously, coherence-enhancing diffusion is an effective method on medical US image processing.This paper mainly researches application of coherence-enhancing diffusion based on PDE in medical US image processing. Firstly, it introduces the research background and the properties in US image. Secondly, coherence-enhancing diffusion filter is discussed, and combined with classical structure tensor, its mathematic models are derived in 2D and 3D respectively, then the corresponding algorithms and experiment results are given. Finally, an improved diffusion approach is proposed, and a new structure tensor integrating the second-order directional derivative information is designed. The two structure tensors can be complementary to each other and provide reliable coherence filter on different structures, which can control the different diffusion stages by using a new defined switch parameter. In order to detect the effect of the improved diffusion approach applying to medical US image, the experiment has been designed, which shows that, the improved diffusion approach is more effective on processing structure in US arthrosis images, meanwhile it improves visualization of arthrosis structure obviously. Undoubtedly experiment results were very promising improved diffusion for an excellent application to clinical diagnosis and operation planning.