基于Retinex理论的X射线医学图像算法的改进与应用

发布时间：2018-10-12 06:49

【摘要】：从伦琴1895年发现了X射线很短的时间内,X射线就得到了广泛的应用,其中也应用于医学影像。X射线的应用促进了医学的发展。在近几年,人们借助计算机可以把不同角度的X射线影像合成三维图像,并由此产生了诸多诊疗手段,进入了数字化、体层成像、三维仿真的重建阶段。随之而来,各种医学影像处理技术也得到了迅速的发展。但是由于人体组织和结构非常复杂,再加上系统设备和环境等因素对X射线不利的影响,最终导致医学影像质量的下降,这主要表现在边缘细节模糊,对比度差,有时还有明显的噪声,在很大程度上影响了医生对疾病的诊断和治疗。因此,除了应用传统的数字图像处理技术,比如直方图处理、空间域和频率域滤波等方法外,还应当尝试新的改进方法。Retinex理论便是其中之一,它的优势在于可以有效的压缩图像动态范围,增强图像的边缘细节,增强图像的亮度,提高图像对比度,改善图像的视觉效果。对于医学影像细节模糊、对比度低、分辨率低、视觉效果差的特点,恰恰十分合适。为此,针对X射线医学影像的特点,论文提出了基于Retinex理论的复合LRA (Logsig cumulative Reintex Algorithm)算法,主要工作如下：首先,分析了传统的图像增强方法,并研究了它们的特点；其次,构建了噪声模型,并应用最新的图像降噪方法对X射线医学影像进行去噪处理；最后,分析并研究了Retinex理论,并且实现了Retinex发展的各阶段算法,利用神经网络中的对数S形LogSig传递函数取代原有多尺度Retinex中的对数函数,并对图像进行动态范围压缩,并由此获得更好的图像亮度调节能力,并在此基础上提出了本文复合的LRA (LogSig cumulative Retinex Algorithm)算法。通过与原有算法进行比对分析,找到了原有Retinex算法针对X射线医学影像应用的不足,并说明了本文算法对于X射线医学影像增强的优势。
[Abstract]:X-ray has been widely used in medical imaging since Roentgen discovered it in 1895. The application of X-ray has promoted the development of medicine. In recent years, with the help of computer, people can synthesize three-dimensional images from different angles of X-ray images, and many diagnostic and therapeutic methods have emerged, which have entered the stage of digitization, tomography and 3D simulation reconstruction. Subsequently, a variety of medical image processing technology has also been rapid development. However, due to the very complex tissue and structure of the human body, coupled with the adverse effects of the system, equipment and environment on the X-ray, the quality of the medical image will eventually decline. This is mainly reflected in the blurring of the edge details and the poor contrast. Sometimes there is obvious noise that greatly affects the doctor's diagnosis and treatment of the disease. Therefore, in addition to the traditional digital image processing techniques, such as histogram processing, spatial and frequency domain filtering, we should also try new and improved methods. Retinex theory is one of them. Its advantages are that it can effectively compress the dynamic range of the image, enhance the edge details of the image, enhance the brightness of the image, improve the contrast of the image, and improve the visual effect of the image. It is very suitable for medical image with fuzzy details, low contrast, low resolution and poor visual effect. Therefore, according to the characteristics of X-ray medical images, a compound LRA (Logsig cumulative Reintex Algorithm) algorithm based on Retinex theory is proposed in this paper. The main work is as follows: firstly, the traditional image enhancement methods are analyzed and their characteristics are studied. Secondly, the noise model is constructed, and the latest image de-noising method is applied to de-noising the X-ray medical image. Finally, the Retinex theory is analyzed and studied, and the algorithms of each stage of the development of Retinex are realized. The logarithmic S-shape LogSig transfer function in neural network is used to replace the logarithmic function in the original multi-scale Retinex, and the image is compressed in dynamic range. On this basis, a composite LRA (LogSig cumulative Retinex Algorithm) algorithm is proposed. By comparing with the original algorithm, the shortcomings of the original Retinex algorithm for X-ray medical image application are found, and the advantages of this algorithm for X-ray medical image enhancement are explained.
【学位授予单位】：首都师范大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：R81;TP391.41

【参考文献】