一种改进的全变分降噪算法在低剂量工业CT重建图像中的应用

Application of an Improved Total Variation Denoising Algorithm in Low-Dose Industrial Computed Tomography Reconstruction Images

针对低剂量CT重建中的噪声抑制问题,文章将全变分降噪模型应用到CT重建的图像投影域,并介绍了ROF模型及能量泛函的建立,以及在低剂量CT图像处理中的应用。提出一种改进的ROF模型的实现方法。首先,利用梯度下降算法对投影图进行迭低降噪处理。其次,使用滤波反投影算法对工业CT图像进行重建。最后,通过CUDA并行运算实现了整个改进算法的过程,提高了算法的运行时间。通过模拟噪声和锂电池快速在线工业CT设备图像处理结果证明:所提出的方法能有效降低重建图像的噪声,提高图像峰值信噪比。

Aiming at the noise suppression problem in low-dose CT reconstruction,the total variation noise reduction model is applied to the image projection domain of CT reconstruction,and the establishment of ROF model and energy func-tional is introduced,as well as its application in low-dose CT image processing.An improved implementation method of ROF model is proposed.First,the gradient descent method is used to perform iteration and noise reduction processing on the projection image.Then,the filtered back-projection algorithm is used to reconstruct the industrial CT image.Finally,the en-tire process of improving the algorithm is implemented through CUDA parallel operation,and the operation of the algorithm is improved time.The results of fast online industrial CT equipment image processing using simulated noise and lithium bat-teries prove that the proposed method can effectively reduce the noise level of the reconstructed image and improve the peak signal-to-noise ratio of the image.