改进数学形态学方法在医学图像处理中的应用

Application of Improved Mathematical Morphological Methods in Medical Image Processing

近年来,随着计算机辅助诊断系统和远程医疗在医学中的快速发展,数字图像处理非常关键。图像的边缘涵盖了大部分的图像信息,医学图像边缘检测是进行后续图像处理的基础。因此,研究医学图像边缘检测具有重要实际意义。基于医学图像成像过程中光源单一以及探测手段的影响,会导致图像噪声分布不均一,往往夹杂多种不同噪声,本文针对医学图像的特性提出了一种改进的形态学算法,包含以下三种优势。第一,自适应权重赋值。对于多方向结构元素,本文算法根据边缘马氏灰度距离自适应赋值各个方向的权重;对于多尺度多形状结构元素,根据信息熵自适应赋值各个结构元素的权重。第二,改进的形态学算子。基于现有算子检测边缘锯齿状、抗噪效果不显著的缺点,本文算法中提出了一种新型抗噪形态学算子。第三,应用于混合噪声彩色医学图像边缘检测。基于现今形态学常应用于灰度图像,为了验证本文算法的鲁棒性,将本文算法应用于四种混合噪声彩色图像进行边缘检测,检测效果良好。最后本文通过视觉直观分析和客观评价指标验证了本文算法均好于其它算法。实验结果表明本文算法提取到的图像边缘完整且清晰,对多种不同混合噪声的抑制和消除也有明显的优势,在医学图像研究中具有很好的应用价值。

In recent years, with the rapid development of computeraided diagnostic systems and telemedicine in medicine, digital image processing is critical. The edges of the image cover most of the image information, and medical image edge detection is the basis for subsequent image processing. Therefore, it is of great practical significance to study the edge detection of medical images. Based on the influence of a single light source and detection methods in the process of medical image imaging, which will lead to uneven image noise distribution, often mixed with a variety of different noises, this paper proposes an improved morphological algorithm for the characteristics of medical images, including the following three advantages. First, adaptive weight assignment. For mul-ti-directional structural elements, the algorithm automatically assigns weights in each direction according to the Marginal Martens grayscale distance, and for multi-scale multi-shape structural elements, adaptively assigns weights to each structural element according to information entropy. Second, improved morphological operators. Based on shortcomings of the existing operator detect-ing edge jaggedness and the noise resistance effect being not significant, a new type of anti-noise morphological operator is proposed in the algorithm. Third, it is applied to the edge detection of mixed noise color medical images. Based on the fact that morphology is often applied to grayscale images, in order to verify the robustness of the proposed algorithm, the proposed algorithm is applied to four mixed noise color images for edge detection, and the detection effect is good. Finally, through visual intuitive analysis and objective evaluation indicators, this paper verifies that the proposed algorithms are better than other algorithms. Experimental results show that the edges of the images extracted by the proposed algorithm are complete and clear, and the suppression and elimination of a variety of different mixed noises also have obvious advantages. It has good application value in medical image research.