Research On B-type Ultrasonic Speckle Denoising Method Based On Speckle Detection

Medical ultrasound is safe, convenient, real-time, low cost and high efficiency, owing to these advantages, it becomes one of the most important diagnostic tool. However, widely in coherent image such as medical ultrasonic image, there is a kind of granular speckle noise, its characterization is pixel gray value changes abruptly, masks and blurs the image details. For the vast majority of clinical diagnosis, details of medical ultrasound images are essential, therefore speckle should be suppressed. In this paper, aiming at the characteristics of the speckle noise in medical ultrasound images, three novel methods for speckle reduction based on speckle detection are presented.Removing speckle need to be aware of the relevant knowledge of the speckle. This article simply introduces the principle of ultrasonic imaging, medical ultrasonic B mode ultrasonic history, the types of ultrasonic imaging, speckle distribution features and the noise model of speckle.First of all, Using local ultrasonic speckle exponent constitute speckle similarity function, and then using the similarity function instead of diffusion function of original anisotropic diffusion. An speckle reduction method based on speckle detection using speckle exponent was put forward. Compared with traditional methods, this method can effectively suppress speckle.The second, in medical ultrasound image, variances of the speckle area changed little, but variances between tissue and speckle region changed a lot. An ultrasound speckle reduction method based on speckle detection using analysis of variance was presented. Variances of bilateral filtering can be adaptive by calculating the gradient of variances of original ultrasonic image, and bilateral filter can filter out speckle on certain degree.Finally, Inspired by the local binary pattern theory, an ultrasound speckle reduction method based on speckle detection using local binary comparison are presented. This method used neighborhood pixels comparative method of local binary pattern. It calculated the pattern values of each pixel, finally filtering out speckle with bilateral filter. The proposed method can obtain better de-speckle and preserve detailed feature of the image.Experiments in phantom testing and in vivo images show that our method can effectively preserve edges and reduce speckle.