Abstract: In order to enhance the quality of the CT image and help to post-process industrial CT image, the method of removing Gaussian noise in the CT imaging process is investigated. In view of the disadvantages of NLM, another method based on the neighborhood window has been proposed. Because calculated similarity of CT projection image using NLM method is inaccuracy, the new similarity has been significantly improved through adding Gaussian window function. This method is compared with the others. The experimental results show that the de-noising algorithm outperforms the others due to the use of neighborhood window. It not only can satisfy the basic needs of de-noising in the process of CT imaging, but also can keep the details of the object better.

Key words: CT, image de-noising, non-local means, neighborhood window

摘要： 为了进一步提高CT图像的质量以利于工业CT图像后处理，对CT成像过程中高斯噪声的去除方法进行了研究。针对NLM算法的不足，提出了一种基于邻域加窗的非局部均值CT成像去噪方法。主要是对CT投影图像数据采用非局部均值技术中相似性计算不准确的问题进行改进，在计算相似性权值时加上高斯窗函数。与其他去噪方法进行对比实验。实验结果表明：采用了邻域加窗非局部均值去噪方法，比其他的去噪方法效果更好。基本满足在CT成像去噪的同时更好地保留工件的细节信息的要求。

关键词: CT, 图像去噪, 非局部均值, 邻域加窗