文档介绍：摘 要
超声成像技术在医学诊断领域获得了广泛应用,相对其它成像技术,它具有成本低廉、使用方便、安全性好,能实时成像等优点。然而,在超声成像过程中,当超声波长与照射物体表面粗糙度相当时,就会产生斑点噪声,这一现象可以用随机散射模型来解释。斑点噪声的存在降低了超声图像质量,掩盖和降低了图像某些细节信息,给后续的图像特征提取和识别、疾病诊断及定量分析造成了不利的影响。
传统的去噪方法包括邻域均值平滑滤波,多幅图像的相干平均,中值滤波,频率域低通滤波器,以及小波去噪等。多幅图像相干平均法需要对同一目标区域不同时刻的多幅图像进行平均,对于肝癌 B 超图像实现起来有一定的条件限制。邻域均值平滑滤波算法简单,但是对原图像易产生严重的模糊,便得边缘的细节信息损失较多。中值滤波法对保留图像边缘以及细节信息有较好的效果。后来,又提出了中值滤波方法和基于偏微分方程的去噪方法。
本文首先运用上述传统方法对肝癌 B 超图像进行斑点噪声抑制实验,比较并分析了这些方法在去除斑点噪声中的优缺点。然后,改进了其中的中值滤波算法,实现了两种新的算法,一是能够判断噪声点的中值滤波算法,二是迭代的中值滤波算法。另外,在小波算法和中值滤波算法的基础上,综合了这两种单一的去噪方法, 实现了小波算法与中值算法相结合的新方法。本文的最后,利用主观及客观评价方法对去噪效果进行评价,得出了如下结论:通过主观上的定性观察以及客观上的定量分析,发现改进的去噪算法比传统算法具有更多的优势,即在去噪的同时更好地保留了图像的边缘和结构信息。
关键词:超声图像;斑点噪声;去噪;小波去噪;中值滤波。
Abstract
Ultrasound Imaging is widely used in the field of medical diagnosis. Compared with other imaging technologies, it has several advantages such as low cost, convenience, good safety and real-time imaging. However, during the diffuse in imaging, speckle noise was created when ultrasound wavelength parable to the size of surface coarseness of the object according to random scattering model. Due to this noise, ultrasound image has low quality, and many detail information of the original image was degraded somehow. As a result, the feature detection, pattern recognition, illness diagnosis and analysis based on the ultrasound image e more difficult than other medical images.
For the part of denoising, many methods including averaging, averaging filtering, median filtering, wavelet soft threshold and PDE method have been proposed. The averaging needs the same target area at different moments of many images to average, which is difficult to obtain for liver cancer B-image. Neighborhood mean smoothing filter is simple, but it will blur the edge causing the loss of edge details. Median filter is good at the reservations of the edge details and has better results. Later, the weighted median filtering methods and the denoising based on partial differential equations are used.
First this thesis us