【作者】 徐艳蕾；

【导师】 赵继印；

【作者基本信息】 吉林大学 ， 通信与信息系统， 2009， 博士

【摘要】 医学图像三维重建方法是医学影像处理系统中的重要组成部分,也是医学图像可视化技术的重点研究问题。本文在深入研究医学CT图像的特点及顺序形态学理论的基础上,主要对基于顺序形态学理论的医学CT图像三维重建方法进行深入研究。针对医学CT图像三维重建方法的图像滤波、边缘检测、分割和三维重建算法等核心内容和关键技术进行研究,主要研究内容如下:(1)医学CT图像滤波平滑算法的研究。在分析顺序形态变换滤波特性的基础上,提出了基于多尺度多结构顺序形态变换的医学CT图像平滑滤波算法,该算法不仅能较好地保持图像的边缘和细节,而且增强了恢复图像的对比度。(2)医学CT图像边缘检测算法的研究。在顺序形态学理论的基础上构建了多尺度多结构元素边缘检测算子,该算子具有提取准确和边缘清晰的特点,而且能够滤除噪声。对于大尺寸图像,顺序形态变换运算速度较慢,因此本文研究了一种快速的基于顺序形态学的医学CT图像边缘检测算法,此算法获得的边缘准确、清晰,而且运算速度大大提高,为图像快速处理奠定了基础。(3)医学CT图像分割算法的研究。通过对顺序形态梯度特性的研究,提出了基于顺序形态梯度的改进的区域增长分割算法和基于顺序形态梯度重构的分水岭分割算法。改进的区域增长算法能够准确分割CT图像中的感兴趣区域,而基于顺序形态梯度重构的分水岭分割算法解决了传统分水岭分割算法的过分割问题。(4)医学CT图像三维插值算法的研究。本文提出了基于顺序形态学的医学CT图像三维插值算法,该算法减小了插值切片间的误差,具有边界清晰、准确和结构清楚的优点,为后续的医学CT图像三维重建奠定了良好的基础。(5)医学CT图像三维重建算法的研究。在基于分割的MC面绘制三维重建算法的基础上,将改进的区域增长分割算法与其相结合,提高了医学CT图像三维重建的精度和运算速度。本文对医学CT图像三维重建的基础内容进行了深入的研究,为医学图像三维重建系统的开发奠定了理论基础,在疾病的诊断和治疗方面具有重要的理论意义和实际应用价值。更多还原

【Abstract】 Computerized Tomographic equipment is a device of illness detection, which has very important meaning in diagnostic medicine.At present, most medical CT equipments can produce digital image,but the doctors diagnose illness by observe slice images produced by a printer or X-film, for the equipment problems.These 2-d slice images are difficult to understand, and it is more difficult to think of 3-d shape and relation of tissue and organ according to these 2-d slice images, so the 2-d image is difficult to meet the demand of medical diagnose and cure. In order to improve the accuracy and scientificity of medical diagnose, it is very important to investigation how to transform the intuitional 3-d medical image from 2-d CT images. The 3-d image can demonstrate 3-d structure and form, and can provide much anatomical information which can’t be obtained by traditional means, and also can provide vision interactive for further simulation. Therefore it has important theory meaning and technology value to study 3-d reconstruction methods of CT images for development of our country medical and health care, and it also can provide theory basis for software platform development and system commercialization of medical image processing, then it has wide application prospect.The 3-d reconstruction technology of medical image not only includes the basic contents such as data acheivement and transform, pre-processing, segmentation and 3-d reconstruction algorithm, but includes design of 3-d visual system and its commerce.Recently, the mathematic morphology has wide application in aspect of computer vision, image processing and analysis, pattern recognition and so on. The mathematic morphology can resolve many problems of image processing, including noise restraint, characteristic extraction, edge detection, image segmentation, shape recongnization, texture analysis, image restore and reconstruction, image compress and so on. At present, mathematic morphology has run to image morphology. In the domain of image morphology, the study of image processing based on order morphology is just the beginning.The paper took up with the study of 3-d reconstruction of CT medical image based on order morphology theory. The main contents of this paper are basic contents including filter, edge detection, segmentation and 3-d reconstruction algorithm of medical image. The following is the detailed study contents:(1) Study on filter of CT image, the smoothness arithmetic based on multi-scale order morphological transformation is proposed.This paper investigates several common filter methods of medical image, and deeply analyzes the mathematic basis and characteristic. The filter property of morphological open and close operation is also studied. Based on this, the property of image’s order morphological transformation and the filter property are studied and analyzed.The order morphological transformation can be considered a filter, which has nature property of filter. The gray-scale image can be made many kinds of processing using different percent values, which can make different transformations and corresponding operation. Then it can realize the filter processing.According to this, this paper proposed the smoothness arithmetic based on multi-scale order morphological transformation. The simulation experiment is done, the result indicates the arithmetic this paper proposed not only can keep the edge and detail of images but can enhance the contrast of images.(2) Study on edge detection of CT image, the edge detection operator of order morphological transform is constructed based on the morphology theory, and the fast arithmetic of edge detection of medical CT images based on the order morphology is also be proposed.This paper studies several common edge detection methods of medical image, and deeply analyzes the mathematic basis and characteristic. The edge detection property of order morphological transformation is studied and analized.When the structure element of order morphological transformation is in smooth area, the output image is almost same to the input for the gray values are nearly equal in this area; but when the structure element is in changing area, the output image fallor rise in the region of gray changing, that is the great differences exist between the input and the output image. Making use of this, the edge can be recognized. The output effect of edge also can be changed by controlling the size of structure element and percent value. Based on this, the paper constructed the edge detection operator based on multi-sacle and multi-structure order morphological transform. The simulation experiment is done and the result indicates the operators can obtain the precise, clear edge of noisy CT images.The edge detection operator of order morphology operated slowly to large-size images. According to this, this paper proposes a fast arithmetic of edge detection of medical CT images based on the order morphology, which can achieves the image edge fast and precisely, especially to the large-size images. By simulation, the arithmetic is proved a fast and precise method and established base for image real-time processing.(3) Study on segmentation of CT images, this paper studies the order morphological gradient image and proposes the region growing arithmetic based on order morphological gradient image. The watershed segmentation of CT images based on order morphology gradient reconstruction is also proposed.This paper studies several common segmentation methods of medical image, and deeply analyzes the mathematic basis and characteristic. The gradient property of order morphological transformation is studied and analized. The gradient map of order morphology is sensitive to the edge of pathological region, so the paper proposed improved region growing arithmetic based on order morphological grade. This arithmetic can segment the interesting region exactly by choosing right seed and float threshold, and the medical affairs can obtain the useful information to diagnose. Furthermore, it provided the precondition for the farther medical image processing, such as extraction of pathological region, measure of specific tissue and 3-D reconstruction.The paper study the characteristic of watershed transform, aiming at the over-segmentation of watershed algorithm, the watershed segmentation of CT images based on order morphology gradient reconstruction is proposed, which can resolve the problem of over-segmentation and has wide application prospect.(4) Study on 3-d interpolation, the interpolation arithmetic of CT images based on percent dilation and erosion operations is proposed.Firstly, the paper analyzes development and shortage of exiting interpolation arithmetic and studies the mathematic mechanism and the macroscopical view of percent dilation and erosion. Based on this, the interpolation arithmetic is proposed. By simulation, the arithmetic is proved an effective method, and the produced slice image is similar to original slice image, which can eatablish well basis for latter 3-d reconstruction.(5) Study on 3-d reconstruction arithmetic of CT images, the improved MC surface rendering algorithm is proposed. At present, there mainly have two 3-d reconstruction methods of medical CT image: surface rendering and volume rendering. The MC algorithm is a common method of surface rendering. MC algorithm can produce clear image to interesting isosuoface and can realize rendering quickly using existing image hardware.The paper mainly studies the MC surface rendering for fast speed and high efficiency.The paper adopts the improved MC algorithm based on segmentation, using the former segmentation images. This algorithm adopts surface track to detect cubes, this can reduce the detection of useless cubes. At the same time, the mthod uses midpoint instead of interpolation, which can improve the precision and speed of 3-d reconstruction.This paper studies the basic 3-d reconstruction contents of CT medical images, which provides theory foundation for exploitation of 3-d reconstruction system of medical images. It has important theory significance and practical application value.更多还原