|
Medical Ultrasound has no harmful , convenient , real-time , etc., in clinical diagnosis, treatment and surgical navigation has a very wide range of applications . However, due to ultrasound imaging coherence properties of the image exist speckle noise that obscures the image of the important details , reduce the body's soft tissues of normal tissue and diseased tissue contrast , these negative factors to the ultrasound image edge detection, image segmentation and subsequent processing such as image registration has caused difficulties. Inhibition of ultrasound speckle noise in image sequences can improve image quality and enhance the image characteristics of diagnosis-related information . Therefore, the study effective ultrasound image denoising algorithm has important theoretical and practical significance . Ultrasound image noise is mainly multiplicative speckle noise associated with the organizational structure information . The classic ultrasound image filtering methods, such as anisotropic diffusion filtering method , the squeeze box filter (SBF), partial differential equations and other methods can not be better speckle noise filtering . Non- local means algorithm is proposed in recent years , a new image filtering algorithm can well protect the image edge and texture information, but this method is running slow, and the filtered image may produce artifacts. This thesis focuses on the above-mentioned drawbacks is proposed based on geometric moments of fast non-local means algorithm . Firstly , proposed based on geometric moments of non-local means algorithm . Geometric moments to modify the original similarity measure, be able to find more similar images . And different orders of geometric moments to do some comparison to find better results can be obtained geometric moments of order. Secondly, the zero- order moment based on the rapid recovery and block non-local means algorithm , greatly reducing the non-local means algorithm running time , and its application to ultrasound image processing . Finally, based on local and non- local means hybrid algorithm , this method can effectively remove multiplicative noise , and applied to the video sequence denoising . Simulation and experimental results of the actual image image evaluation parameters (PSNR, MSSIM, ENL) show that the proposed algorithm can effectively filter the speckle noise , protect the image edge information and further details.
|