论文部分内容阅读
在小动物计算机断层扫描(CT)实验中,因需考虑小动物存活率以及实验的连续性等问题,一般较少采用高剂量的X射线进行实验;而低剂量的X射线会导致重建图像被噪声污染,影响图像质量,不利于后续实验分析。为解决此问题,本文介绍了一种基于全局字典学习的降噪方法,并将其应用于提升低剂量小动物CT重建图像质量的研究中。针对真实的小动物CT重建数据,选择高剂量的小动物CT重建图像作为训练样本,利用逐列更新的字典学习算法(K-SVD),构建包含图像信息的全局字典;利用正交匹配追踪算法(OMP)将低剂量重建图像利用全局字典进行稀疏分解,分离噪声,最后将重建图像复原,达到降噪、提升图像质量、降低小动物CT实验的拍摄剂量、提高小动物存活率的目的。实验结果表明,本文提出的方法能够有效减少低剂量动物CT图像的噪声,并能够较好地保留图像细节。
In small animal computed tomography (CT) experiments, small doses of X-rays are generally less commonly used for experimentation because of concerns such as small animal survival and experimental continuity; low dose X-rays can result in reconstructed images Noise pollution, affecting the image quality is not conducive to follow-up experimental analysis. In order to solve this problem, this paper introduces a noise reduction method based on global dictionary learning, and applies it to enhance the image quality of CT reconstruction of low dose animals. Aiming at the real CT reconstruction data of small animals, high-dose CT reconstruction images of small animals were selected as training samples, and a global dictionary containing image information was constructed by using dictionary learning algorithm (K-SVD) updated column by column. Using orthogonal matching pursuit algorithm (OMP), the low-dose reconstructed image is sparsely decomposed using a global dictionary to separate the noise. Finally, the reconstructed image is reconstructed to achieve the purpose of reducing noise, improving the image quality, reducing the shooting dose of the small animal CT experiment and improving the survival rate of the small animal. The experimental results show that the proposed method can effectively reduce the noise of CT images in low-dose animals and preserve the image detail well.