中文摘要：

在分析航空发动机涡轮叶片工业计算机层析成像（computed tomography,CT）图像自相似性特点基础上,讨论了一种基于图像不同区域相似性的非局部平均降噪算法,首先采用高斯加权欧式距离算子计算同一图像中不同区域间的相似性,然后对具有较强相似性的区域进行叠加平均,从而降低噪声,同时保持图像边缘的对比度.针对非局部平均算法计算复杂度高、计算量大的问题,研究了一种基于傅里叶变换的非局部平均降噪加速算法,以欧式距离算子替代高斯加权欧式算子提高相似性的计算效率,减少计算时间.实验结果表明,在保证航空发动机涡轮叶片工业CT图像降噪效果的情况下,处理速度提高了4倍以上.

英文摘要：

The similarity between different regions in an X-ray industrial computed tomography（CT） image of aeronautical engine turbine blade has been analyzed.A method called non-local means algorithm based on the similarity has been discussed to improve the industrial CT image contrast.This method adopts the weighted-Gaussian Euclidean distance to calculate the similarity between different regions in an image.Then it averages the regions with the similarity.It can decrease the noise in industrial CT images and meanwhile reserve the edge sharpness.However this method has a high computational complexity.So an acceleration algorithm based on Fourier transform for the non-local means algorithm has been researched.The weighted-Gaussian Euclidean distance is also replaced by the Euclidean distance to improve the calculation efficiency.The experiment results show that,this algorithm can improve the speed at least four times without affecting the denoising effect.