工业CT图像环形伪影校正

为了去除工业CT图像中的环形伪影,提高CT重建图像质量以及后续处理和量化分析的精度,提出了一种基于投影正弦图的新型校正方法。首先,采用S-L滤波器对原始投影数据进行滤波,增强伪影信息。接着,对滤波后的投影数据进行线积分,并采用差分处理,进一步增大伪影与工件轮廓的差异。然后,按照插值次数对差分后的投影数据进行插值平均,并根据正态分布选取自动查找伪影的位置。最后,结合线性插值与线性外推的方法对环形伪影处投影数据进行校正。对含有环形伪影的实际CT图像进行了校正实验,结果表明,与基于极坐标变换、小波-FFT滤波的校正方法相比,该方法校正后的灰度图像信噪比增益达1.688 dB,有效地消除了环形伪影,同时又很好地保持了图像边缘及分辨率。对探测器故障或性能不稳定引起的间断的环形伪影校正效果良好。

Ring artifact correction for ICT images

A new image correction method based on projection sinograms is proposed to remove ring artifacts in industrial Computerized Tomography (CT) and to improve the precisions of post processing and quantitative analysis. An S-L filter is adopted to filter original projection data and to enhance artifact information. Then, the filtered projection data are processed by line integral and difference to further increase the difference between artifact edges and work piece contours. Furthermore, the differenced projection data are averaged according to the number of interpolations, and the artifact location is automatically searched according to the threshold selected by normal distribution. Finally, the projection data of ring artifacts are corrected by combining the linear interpolation and linear extrapolation. A correction experiment is carried out in the actual CT images including ring artifacts. The result indicates that the SNR of gray image using the proposed method has increased by 1.688 dB, which shows that proposed method effectively eliminates the ring artifacts and keeps the image edges and resolution as compared with the methods of polar coordinate transformation or Wave-FFT. The proposed method is suitable for correcting the discontinuous ring artifacts caused by detector faults and unstable performance.