【作者】 苏曦；

【导师】 孙杰；

【作者基本信息】 天津理工大学 ， 光学工程， 2007， 硕士

【摘要】 随着计算机技术以及仪器数字化技术的发展与普及,数字化医疗技术也得到了相应的促进与提高,计算机化X线放射成像系统即CR(Computerized Radiography)成像系统就是其中具有代表性的数字化医疗技术之一。CR影像系统在成像过程中不可避免地会混入各种噪声,引起图像质量的下降,进而影响以图像为判别依据的医学诊断正确率。因此,要提高医学影像判读的正确率就必须通过对医学影像进行滤波处理来降低或消除图像噪声。在具体实现图像噪声滤波操作之前,有必要认清干扰图像信息的各种噪声来源、特征及其与有用信号的关系。由于图像噪声滤波的针对性很强,所以不同滤波方法对不同噪声的处理效果的选择特异性是有效运用数字图像处理技术来提高CR影像质量的关键。小波变换是以傅立叶变换为基础发展起来的数学分支。小波分析将时域和频域结合起来描述信号的时频联合特征,并且具有在时频两域表征信号局部特征的能力。目前,小波分析已广泛应用在语音识别、图像处理、图形分析、通信设计、地震预报、生物医学工程、计算机视觉等领域并取得的成功,为将小波分析方法用于CR影像噪声滤波提供了可行性条件。本文重点研究了CR影像噪声滤除的小波分析方法,主要工作包括:1.系统介绍了CR成像系统的设备结构与工作原理;分析了CR成像系统的成像过程及图像特点;总结分析了CR影像的噪声来源;通过系统分析得出主要的两种噪声分别服从加性高斯分布和乘性泊松分布;深入讨论了两种主要噪声与图像有用信号之间的关系及在一定范围内对噪声进行识别的方法。2.介绍了小波变换的基本理论,特别是分析了小波变换多分辨率分析的特点,这些理论分析为基于小波变换滤波方法的提出奠定了基础。3.分别详细讨论了经典图像滤波方法与四种小波滤波方法的原理和算法,以加性噪声为例,理论分析和滤波实践证明了基于小波变换的滤波方法优于经典滤波方法的结论;并分析了各种方法的优缺点及适用条件;并对其中的模极大值滤波方法的算法实现进行了分析;通过对四种小波变换滤波方法的进一步分析与比较,提出了改进并简化模极大值小波变换去噪方法的运算算法;得出小波变换系数域相关性滤波方法和改进后的小波变换模极大值的滤波方法为最优图像噪声滤波方法的结论,并且得出该方法适用性很强的结论;给出不同图像噪声滤波方法的适用条件。4.在对具有高斯噪声和脉冲噪声的图像噪声滤波分析的基础上,对CR影像中的乘性泊松噪声处理也进行了分析比较,给出了泊松噪声滤波效果的分析,各种方法比较后,得出了中值滤波方法对于乘性泊松噪声的滤波效果最优,其次是基于小波变换的强制去噪方法。更多还原

【Abstract】 With the development of computer technology and digital technology, digitized medicine technology has developed, while CR (computed radiography) is one of the typical methods.In the process of computed radiography, it is inevitable to bring many kinds of noise in the computed radiography images. So the images have been degraded, which can worsen the result and the analysis of doctors. So in order to improve the validity of the computed radiography images it is necessary to filter the noises. And before the filtering, we must firstly get cross where the noises come from and what the noises’character is and what kinds of relationship between the noise and the useful information from the images. Then we can do some pertinent filtering. Since different methods will have different effects of the filtering, we should base on the specialists of computed radiography and digital image processing to study filtering methods for computed radiography images noises.Wavelet transform is a new embranchment of mathematics based on Fourier‘s transform. It can combine time domain and frequency domain to describe the character of time-frequency of signals, and it also can depict the local feature of the signals in the time domain and frequency domain. At present, wavelet transform is widely used in voices, images, graphics, communication, earthquake, computer vision and so on. So it also can be useful in filtering the noises of CR images. Our major work:i. Firstly we introduce the equipment and the theory of the computer radiography system, analyses its character and imaging process, based on these, we know the source of the noises and the major two kinds of noises are Gaussian noise and Poisson noise. Then we discuss the relationship between the noises and the useful information in the images and distinguish them in some extension.ii. We introduce the basic theory of the wavelet transformation especially the Multi- resolution Analysis (MRA), These theories are bases of the filtering methods using wavelet transformation.iii. We analyze four different filtering methods using wavelet transformation and develop the method of using modulus maximum to make it facility. Then we compare these methods each other and with classical filtering methods. From these we can put forward that methods based on wavelet transform are better than classical filtering methods and the methods of modulus maximum and scale relativity are better. At last, we bring forward the application of methods.iv. Above all, we use all these methods to filter the Gaussian noise and Poisson noise. Then we use these methods on Poisson noise. Through the comparison and the analysis, we find that wavelet transformation filtering methods are still useful for images with Poisson noise, but the medial filtering is the best method for multiplicative Poisson noise.更多还原