光学学报, 2018, 38 (4): 0410003, 网络出版: 2018-07-10
基于卷积神经网络的低剂量CT图像去噪方法 下载: 1373次
Low-Dose CT Image Denoising Method Based on Convolutional Neural Network
图像处理 图像去噪 低剂量计算机断层扫描 深度学习 卷积神经网络 image processing image denoising low-dose computed tomography deep learning convolutional neural network
摘要
为了改善低剂量计算机断层扫描(CT)图像的视觉质量,提出一种基于卷积神经网络的图像去噪方法。网络引入批量归一化,并且学习的是低剂量CT图像到其噪声图像之间的映射;使用空洞卷积在不提高复杂度的情况下增大感受野;此外,还将前后层的特征图进行连接,使后方的卷积层能够利用前方各层的特征图作为输入,鼓励网络中特征图的重用。实验结果表明,与目前较先进的方法相比,所提网络结构在实现了更好去噪效果的同时大幅度降低了网络复杂度,能够快速、显著地改善低剂量CT图像的视觉质量。
Abstract
In order to improve the visual quality of low-dose computed tomography (CT) images, an image denoising method based on convolutional neural network is proposed. The batch normalization is introduced to the network, and the mapping function between low-dose CT images and their corresponding noise images is learned. The dilated convolution is used to expand the receptive field without increasing the complexity. Besides, the feature maps from the front and back layers are concatenated, and all the feature maps of convolution layers ahead can be used as the input of a subsequent convolution layer.It encourages the reuse of feature maps in the network. The experimental results show that the proposed network architecture achieves better denoising performance and sharply reduces the network complexity when compared with the state-of-the-art method at present. So, it can quickly and significantly improve the visual quality of low-dose CT images.
章云港, 易本顺, 吴晨玥, 冯雨. 基于卷积神经网络的低剂量CT图像去噪方法[J]. 光学学报, 2018, 38(4): 0410003. Yungang Zhang, Benshun Yi, Chenyue Wu, Yu Feng. Low-Dose CT Image Denoising Method Based on Convolutional Neural Network[J]. Acta Optica Sinica, 2018, 38(4): 0410003.